[ Ilias Tsitsimpis ]
* Change Priority to optional. Since Debian Policy version 4.0.1,
priority extra has been deprecated.
* Use the HTTPS form of the copyright-format URL
* Modify d/watch and Source field in d/copyright to use HTTPS
* Declare compliance with Debian policy 4.1.1
[ Sean Whitton ]
* Add fix-hssplicety.patch (Closes: #867195).
[ John Paul Adrian Glaubitz ]
* Make ghc cross-buildable
[ Gianfranco Costamagna ]
* Drop arm64 switch to gold, now gcc/binutils should be fixed.
* Team upload.
* Team upload.
* Patch arm64 before autoreconf.
* Team upload
* Try arm64 with my Ubuntu patch.
* Only apply arm64 linker patch on arm64, to avoid conflict with
armel-only patch.
* Switch arm64 linker to bfd, since gold is garbage.
* Patch from James Clarke to fix powerpc breakage caused by
powerpcspe patch.
* Revert cross-compilation patch.
[ John Paul Adrian Glaubitz ]
* Build with --enable-unregisterised on powerpcspe. closes: #861806.
* Support building ghc without building haddock, for cross-compilation.
closes: #853285, #853773.
[ Helmut Grohne ]
* Do not run tests under DEB_BUILD_OPTIONS=nocheck (Closes: #853743).
[ Clint Adams ]
* Conflict with libghc-cabal-dev (<< 1.24.2.1) to avoid wanna-build
confusion.
* Team upload
* New upstream release, remove patches addressed upstream
-add-missing-MO_WriteBarrier
-no-pie
-find-tycon-panic
-get-linker-flags-correctly
-fix-ppc-lwa-generation
-smp-arm-fix.patch
-R_X86_64_REX_GOTPCRELX
* Remove old prof_scc.png link on ghc-doc (Closes: #855627)
* Bump std-version to 4.0.0
* Bump compat level to 10, dropping autoreconf
* Add upstream patch "Add relocation type R_X86_64_REX_GOTPCRELX".
* Patch to fix SMP support enablement on armhf.
* Add osdecommitmemory-compat.patch (Closes: #847677).
* Upload to unstable.
* Include ghc --info output in build log.
* PPC/CodeGen: fix lwa instruction generation. closes: #844323
* Use upstream patch for '-no-pie'.
Drop our custom 'opt-pic' and 'no-pie' patches and use one provided by
the upstream developers.
* Team upload.
[ John Paul Adrian Glaubitz ]
* Add patch x32-use-native-x86_64-insn.patch to support
x32 architecture (Closes: #814657)
[ Emilio Pozuelo Monfort ]
* Re-enable parallel builds, broken since 7.10.1.20150630-1.
[ Clint Adams ]
* Upstream patch c30b179a73d9fd3f6edcdda5e881523cd6edd46a to Cabal,
fixing incorrect retrieval of ld flags.
[ Ilias Tsitsimpis ]
* Add upstream patch to avoid find_tycon panic (Closes: #842679).
[ Clint Adams ]
* Patch from James Clarke to prevent unaligned accesses on sparc64.
closes: #842780.
* Reduce memory usage when building on mips/mipsel.
* Upload to unstable to start GHC 8 transition.
* Disable PIC on powerpc.
* Patch GHC to pass `-no-pie' along with `-Wl,-r'.
* Do not disable PIE during package configuration in d/rules.
The above patch seems sufficient.
* Switch PIC default to on for all architectures.
[ Balint Reczey ]
* Disable PIE to fix build when PIE is the default in gcc
(change taken/revised from Ubuntu, Closes: #712228)
* Removed the empty ghc-dbgsym package.
[ Joachim Breitner ]
* Merge from unstable (up to 7.0.3-9)
[ Matthias Klose ]
* Disable PIE on amd64 and ppc64el.
[ Gianfranco Costamagna ]
* Fix VCS fields
[ Sven Bartscher ]
* Depend on haskell-devscripts-minimal instead of duplicating its
functionality.
* New upstream release
* New upstream release candidate (8.0.1-rc4)
* New upstream release candidate (8.0.1-rc3)
* Build-Conflict: Do not bootstrap with ghc-8.0.1-rc1, as this is broken.
[ Gianfranco Costamagna ]
* Re-enable and refresh patches/armel-revert-ghci-fixes.patch on armel.
- drop a little part on Linker.c, upstreamed.
[ Joachim Breitner ]
* New upstream release candidate (8.0.1-rc2)
* Ah, there is llvm-3.7 in Debian. Must have been confused by the LLVM
packaging... let’s use that.
* Try to see if it builds without
patches/armel-revert-ghci-fixes.patch on armel.
* Try to build with LLVM 3.8 (upstream wants 3.7, but that is not in Debian
any more)
* New upstream release candiate (8.0.1-rc1)
Dropped patches
+ cabal-show-detail-direct
+ sh4-platform-detection-support
+ parc64-initial-platform-support
* Configure haddock to use locally provided mathjax files, and let ghc-doc
depend on libjs-mathjax.
* Backport upstream patch D1782 to unbreak manpage building
* Add gpg key ot debian/upstream and use it in debian/watch
[ Clint Adams ]
* Backport upstream 563a4857abcee4a6e43c68323274309c58f42aa0 to use gcc's
SMP primitives instead of GHC's broken ones.
* Backport upstream 7053019e7b04842dd7364039381d8c4c069489a2 to add a
missing MO_WriteBarrier to emitPrimOp.
* Improve debian/watch file.
[ Dmitry Bogatov ]
* Use secure (https) uri in Vcs-Git field in 'debian/control'
* Bump standards version to 3.9.8 (no changes needed)
* Convert `debian/copyright' to dep5 format
[ Joachim Breitner ]
* Ensure that the shebang in the generated wrapper scripts use /bin/sh, and
not the value of SHELL of the current user.
[ Joachim Breitner ]
* Tighten build-depends on llvm-3.5, to pull in backported version
* Tighten build-depends on ghc; this does not build against 7.6 any more, so
let's use the 7.8 already in backports.
[ Colin Watson ]
* Run /usr/lib/ghc-doc/gen_contents_index from ghc-doc.postinst with HOME
set to a temporary directory, so that it doesn't pick up stray user
configuration and so that it works reliably in autopkgtests.
* Let GHC break cabal-install (<< 1.22), due to
"ghc no longer supports single-file style package databases"
* ghc-doc Breaks/Replaces libghc-terminfo-doc (<< 0.3.2.5-4),
libghc-transformers-doc (<< 0.3.0.0-6), libghc-xhtml-doc (<<
3000.2.1-5), now that these are all shipped with ghc (closes: #800423).
* Non-maintainer upload.
* Add debian/patches/armel-revert-ghci-fixes.patch which reverts
the ARM ghci fixes, but apply the patch on armel only with
the help of the debian/rules file (Closes: #807020).
[ Colin Watson ]
* Disable PIE on s390x, since Ubuntu's toolchain defaults to enabling it
there (closes: #807366).
[ Joachim Breitner ]
* Add patch for Sparc64. Again thanks to John Paul Adrian Glaubitz for the
patch (Closes: #807777).
* Platform detection support for sh4 (Closes: #807108). Thanks to John Paul
Adrian Glaubitz for the patch.
* Switch Vcs-Git/Vcs-Browser headers to new location.
* New upstream version.
* Apply a patch by Peter Trommler to fix
"PPC.Ppr: Shift by 32 bits is not allowed." on powerpc
* Patch Cabal to provide --show-detail=direct, mostly to work around
problems with `setup test` on architectures with insufficient threading
support in GHC: https://github.com/haskell/cabal/issues/2398
* New upstream release (7.10.3-rc1)
* Backport patch to fix GHC bug #10549, to unbreak shake’s test suite
* New upstream release
* Remove PPC-relocations.patch
[ Joachim Breitner ]
* New upstream release candidate, 7.10.2-rc2
* Clean up debian/control a bit, based on a patch by Gianfranco
[ Gianfranco Costamagna ]
* Bump std version
* Refactor rules file to use dh
* Calculate haddock interface version using the built haddock, not the
installed (Thanks to David Fox for noticing).
* New upstream release randidate (7.10.2-rc1)
* Merge from unstable
* Merge from unstable (Closes: #785282, #785194 )
* A lucky guess at making GHC build reproducible, by not putting the process
id in temporary filenames.
* Fix patch/ghc-7.8.4-3-aarch64.patch; dropped a hunk when upgrading to
7.10 that I should not have dropped. Thanks Gianfranco Costamagna for
noticing.
* Apply patch by Sergei Trofimovich to implement load_/store_load_barrier on
armv6 and older (upstream ticket #10244)
* Build against llvm-3.5 (Closes: 784245)
* New upstream release. Optimistically dropping lots of patches, lets see
what now works out of the box.
* Ship haddock with ghc, and leave (most) paths as they are. It is easier to
not fight against the way upstream builds and installs stuff.
* Remove the hack introduced in 7.2.0 for a clean upgrade from 7.0.3.
* Another stab at making GHC more reproducible: Use a hash of the command
line instead of the pid when calculating a "random" directory name.
* debian/patches/stable-specialization-rule-names: Root out another cause of
binary ABI instability.
* Remove useless *.haddock.t files (upstream bug 10410)
* A lucky guess at making GHC build reproducible, by not putting the process
id in temporary filenames.
* Hack to get stable ABIs even when the build path changes. Closes:
#785282
* New patch patches/PPC-relocations.patch by Colin Watson, Closes: #785194
* Apply hurd compatibility patch by Pino
* Put .haddock files of libraries shipped with GHC into
/usr/lib/ghc-doc/haddock/ghc/<pkg> instead of
/usr/lib/ghc-doc/haddock/<pkg>. This avoids overriding such a file if a
-doc package of the same package and version is installed.
Closes: #783863.
* Use ld.gold on arm64 and enable GHCi. Thanks to Edmund Grimley Evans for
digging up the relevant upstream patches, and for Erik de Castro Lopo for
writing them in the first place. Closes: #783987
* Remove LLVM path hacks, ghc-7.8 properly remembers the value passed via
--with-opt=opt-3.4 and use it at runtime (at least I think it does).
* Fix debian/rules syntax (tab instead of spaces) to unbreak build on armel,
armhf
* Apply all hacks for armel and armhf also on arm64
* Upload to unstable.
* Fix version glitch now, so that it never affects anything besides
experimental.
* ghc-doc: Conflict with libghc-xhtml-doc
* Merge from unstable:
- Let ghc-haddock break on ghc-doc (<< 7.6.3-20~). Closes #781649.
- Improve watch file
- Mark all triggers -noawait. Closes: #769554
- Joey Hess: armel,armhf: Force use of llvm-3.4, rather than unversioned
lvm, which can have abi changes that break ghc. Closes: #763078
* New upstream release (7.8.4)
* Drop patches always_build_arm_spinlocks.patch, linker-detection-fix,
dll-split-fix.patch, TcSplice-no-GHCi-fix; applied upstream.
* Now that the package builds on armel and armhf, try to build with GHCi
enabled.
* Apply patches taken from GHC HEAD that help building on ARM
- always_build_arm_spinlocks.patch
- dll-split-fix.patch
- linker-detection-fix
* Replace patches/Fix-documentation-build-failure-without-GHCi.patch by the
equivalent patch from GHC HEAD (./patches/TcSplice-no-GHCi-fix)
* Pass -fuse-ld=gold to gcc when used as alinker on arm
(patches/use-gold-on-arm) and make sure this setting does not affect other
parts (patches/saner-linker-opt-handling).
* Try to handle autoreconf saner, by simply calling perl bool
* According to Ben Gamari, the flag is -optl-B..., not -optc-B...
* Do not pass -mlong-calls to gcc on arm, on the grounds that upstream is
not doing it.
* Try using ld.gold on arm, using -optc-B/usr/bin/ld.gold
* Try using ld.gold on arm.
* Explicitly call opt-3.4, and avoid the generic opt command.
* Explicitly depend on llvm-3.4 on arm, maybe that works better.
* New upstream release candidate (7.8.4-rc1)
* Add support for mips and mipsel.
Add mips-support.patch.
Patch by Dejan Latinovic <Dejan.Latinovic@imgtec.com>.
Closes: #751479.
* Depend on libncurses5-dev, since we ship terminfo in ghc now. Thanks to
David Fox for noticing.
* Also ship terminfo, haskeline, and xhtml. Thanks to David Fox for the
patch and testing.
* New upstream release (7.8.3)
The version number is a work-around for a glitch earlier, see below.
* Merge 7.6.3-13 from unstable
* Merge 7.6.3-9 to 7.6.3-12 from unstable
* Do not pass -optc-mminimal-toc on ppc64 (Closes: #747089), thanks to
Hiroyuki Yamamoto for the patch.
* Try building without hardening and see what architectures work then.
* New upstream release (7.8.2)
The version number is a work-around for a glitch earlier, see below.
* Fix haddock build on armhf (Upstream bug #8988, patch by Colin Watson)
* debian/watch: Refer to xz-compressed tarball
* New upstream release (7.8.1)
The version number is a work-around for a glitch earlier. When uploading
to Debian unstable, the proper 7.8.1 can be used. This way, the glich will
be confined to experimental, and only until 7.9 or 7.10 is uploaded to
experimental.
* New release candiate (7.8 RC 2)
* Do not provide stm or parallel, these are not officially part of the GHC
distribution.
* transformers now comes with ghc, put it in the Provides line
* Call ghc and ghc-pkg via the wrappers in /usr/bin in the postinst script
(Closes: 738183)
* New upstream release candidate
* Drop ghc-dynamic, now contained in ghc
* Fix watch file.
* Let ghc-haddock break on ghc-doc (<< 7.6.3-20~). If I used piuparts
correctly, this finally closes: #781649.
* Mark all triggers -noawait. Possibly Closes: #769554
* Fix armel/armhf script munging; ghci and runghc are not available on
those arches.
* Fix accidental line wrap in debian/rules
* armel,armhf: Force use of llvm-3.4, rather than unversioned lvm,
which can have abi changes that break ghc. Closes: #763078
* Also do not conflict with libghc-binary-doc
* Do not conflict with libghc-binary-dev.
* Backport from 7.8: Do not panic when annotations are encountered on non-TH
architectures. Less work this way than patching all the library code.
(Closes: 752460). Thanks to Dejan Latinovic to provide the patch.
* Actually use the conflicting-devs variable in debian/control. No idea what
went wrong yesterday.
* Last update broke the calculation of the Provides fields, fixing.
(Closes: #751051)
* Do no conflict with libghc-cabal-dev.
* Team upload.
* Add ppc64el support.
* Team upload.
* Ensure that all copies of config.guess and config.sub in the tree are up
to date at build time.
* Add arm64 support.
* Apply a4b1a435 from upstream, to fix building on 64 big endian platforms
such as s390x. Thanks to Aurelien Jarno for pointing out the fix.
Closes: #740144
[ Gianfranco Costamagna ]
* Bumped standard version to 3.9.5, no changes required.
[ Joachim Breitner ]
* Depend on gcc (Closes: #736472)
* Closes: #731597: llvm-3.3 compatibility (patch from upstream)
[ Gianfranco Costamagna ]
* Removing gcc from control file build-deps
[ Louis Bettens ]
* fix cabal bug #1087 in current version
[ Colin Watson ]
* Enable verbose build output (see
http://lists.debian.org/debian-devel/2013/06/msg00539.html).
[ Gianfranco Costamagna ]
* Switch to llvm (Closes: #711948)
* removed deprecated DM-Upload-Allowed
* removed some version checks, higher versions are already in oldstable.
* Added dpkg-buildflags as build-dep, fixing lintian warning.
* Add patches/Handle-sign-bit-when-generating-veneer-for-ARM-Thumb.patch,
patch by Colin Watson
* Disable GHCi on arm, as it is severly borken (for more information see
http://hackage.haskell.org/trac/ghc/ticket/7794)
* Do not create system-wide documentation index if /usr/share/doc is
missing. (Closes: #709911)
* Create a symlink from prof_scc.png to prof_scc (Closes: #664043)
* Enable compat level 9
* Remove transitional ghc6-* packages
* New upstream release
* New upstream release
* Typo in ghc-haddock’s Provides, thanks to David Fox for noticing.
* New upstream release (aka 7.6.1-rc).
+ Closes: #677096 (huge number of wakeups)
* Remove debian/patches/armhf_llvm_abi, applied upstream, thanks to Shawn
Landden for noticing (Closes: #695739)
* Hardcode in debian/rules that haddock supportin interface version 22 also
supports interface version 21.
* Bump standards version, no change
* Change VCS fields back to regular repo
* Build the runtime also in the thr_debug_p way, suggestion by Michał J.
Gajda
* Do not Provide transformers, it is only used internally.
* Ensure this is not used with unstable’s haskell-devscripts
* New upstream release
Dropped patches, applied upstream:
+ Fix-GHCi-segfault-on-linux-powerpc
+ hurd-is-ELF
+ use-llvm-3.0
+ memcpy-ffi
+ fix-PPC-right-shift-bug
+ no-useless-time
+ fix-ppc-ghci-segfault
* Configure the use of llc-3.0 and opt-3.0 via ./configure
* Bump standards version
[ Erik de Castro Lopo ]
* Add patch to fix GHCi segfault on startup on PowerPC (Closes: #631073).
* Remove patch no_ghci_on_powerpc.
[ Iain Lane ]
* Update Vcs-* to point to experimental branch
* New upstream.
* debian/patches/series
- Remove patch fix-ARM-s-StgCRun-clobbered-register-list-for-both-A,
fix-ARM-StgCRun-to-not-save-and-restore-r11-fp-regis and memcpy-ffi.patch
(in upstream).
- Refresh other patches.
[ Erik de Castro Lopo ]
* Add debian/patches/fix-PPC-right-shift-bug which fixes upstream GHC bug:
http://hackage.haskell.org/trac/ghc/ticket/6156 (Closes: #677591)
* Refresh other patches.
* debian/contol: Add myself to uploaders and set DM-Upload-Allowed to yes.
[ Joachim Breitner ]
* debian/patches/no-useless-timer: Backported from GHC 7.4.2, (Closes:
#677096)
* Make sure GHC is using ld.bfd, as it passes arguments not understood by
the gold linker (Closes: #673081)
* Use a saner priority for the runhaskell alternative. (Closes: #676970).
TODO for later: Consider removing the alternatives code altogether.
* Remove the ghc-doc symlink logic from 7.0.4-3, as with 7.4.1, every
documentation package had to be rebuit.
* ghc-doc: Conflict on libghc-haddock-doc, as it contains files that moved
into ghc-doc. (Closes: #659081)
* Added debian/patches/memcpy-ffi.patch, backported from upstream master
branch. See http://hackage.haskell.org/trac/ghc/ticket/5967.
[ Iain Lane ]
* Two new patches (backported by Iulian Udrea) to fix armel build
failures. See upstream bug #5824.
- fix-ARM-s-StgCRun-clobbered-register-list-for-both-A
- fix-ARM-StgCRun-to-not-save-and-restore-r11-fp-regis
* Use dh_autoreconf{,_clean} to autoreconf, mainly so we can have proper
clean support.
* armhf support (thanks to Jani Monoses):
- debian/patches/ARM-VFPv3D16: Use vfp3-d16 FPU for ARM builds.
- debian/patches/armhf_llvm_abi: Pass -float-abi=hard to llc on armhf if
__ARM_PCS_VFP is defined (needs to be preprocessed for this)
- debian/rules: Define __ARM_PCS_VFP on armhf for the above patch.
[ Joachim Breitner]
* patches/hurd-is-ELF added, Closes: #659530, thanks to Samuel
Thibault
* Also conflict with any provided libghc-*-dev packages. Thus, if a packages
is moved _into_ ghc that was outside before, no external package can be
installed (as was the case with binary).
* patches/no-missing-haddock-file-warning: Quench warning, as it is quite
common on Debian installations.
* build-depends-indep on hscolour and fop, to ensure most documentation is
built.
* New upstream release.
* New upstream release candidate (7.4.1-rc2)
* s/armef/armhf/ in debian/control
* New upstream release, drop all patches introduced in 7.4.0.20111219-3,
applied upstream.
* Prefix haddock interface number with "haddock-interface" in
Depends/Provides pair
* Switch to llvm and registered builds on arm*
* Do not set UseLibFFIForAdjustors in debian/rules, as the code in
./mk/config.mk.in seems to do the same anyways.
* Do not set GhcWithInterpreter=YES for kfreebsd-*, done in
./mk/config.mk.in by now.
* Do not set GhcUnregisterised=YES in debian/rules, ./mk/config.mk.in
already makes the right choice.
* Do not build the accidentally included dph libraries.
* Explicitly depend on llv-3.0 and use corresponding binary names.
* patches/kfreebsd-OS: KFreeBSD is an ELF-based OK
* patches/S390-Arch: Mention bitness of S390(x)
* patches/ArchMips: Add more missing Arch values
* Fix building bundeled haddock out-of-tree (We need to re-think this
approach).
* Drop patches/lpthread-bootstrap-workaround, probably not needed any
more. Also drop autoconf dependency.
* New upstream release (AKA 7.4.1-rc1)
+ Closes: #635587 (GHCi linker could not find libstdc++)
+ Closes: #514720 (cabal installation wrecks permissions) (was actually
closed earlier, I think)
+ Closes: #533022 (Data.Word should define instances of Random for each
type)
* New upstream release
* Call autoconf and remove configure in clean, to avoid having an
autoreconf-patch in debian/patches.
* Drop patches/hash-version-number, applied upstream
* Drop improve_linker_script_handling, applied upstream
* New upstream release.
* Remove ghc-doc dependency on ghc, it did not have the desired effect.
Instead call ghc-pkg recache in ghc-doc’s trigger as well.
* Install 7.2.1-rc1 to experimental.
* Undo the symlink-workarounds introduced in 7.0.4-3. The -doc work-around
is still needed, as the haddock interface version did not change.
* Suggest llvm package
* Remove ghci support on powerpc (Closes: #631073)
[ Joachim Breitner ]
* patches/configure-s390x: Add s390x to the list of known architectures.
* add armhf in debian/rules (Patch by Riku Voipio, Closes: #640811)
[ Iain Lane ]
* Handle linker scripts for which dlopen returns "file too short" and/or
which contain INPUT statements (for example libncurses.so) (Closes:
#644591)
* Standards-Version → 3.9.2, no changes required
* Fix #609445 by actually removing the patch for #566331, not needed any
more since this upstream commit:
Sat Jul 31 12:55:06 BST 2010 Ian Lynagh <igloo@earth.li>
* Expose the functions haddock needs even when haddock is disabled; #3558
* Remove ghc-doc dependency on ghc, it did not have the desired effect.
Instead call ghc-pkg recache in ghc-doc’s trigger as well.
* Fix missing instance documentation (Closes: #609445)
This was a Debian-specific regression, introduced in the fix for #566331
or later caused by it, and fixed by building haddock with -DGHCI.
* Let ghc-doc depend on ghc. It already does via ghc-haddock, making this
explicit ensures that the triggers run in the right order.
* debian/patches/hash-version-number: Include the upstream version number in
the ABI hash. We do need to recompile everything, even after a minor ghc
upgrade. This allows to use the same mechanism to ensure correctness and
detect that we need to recompile something as for other kind of
recompilations. (Closes: #633828)
* Employ similar symlink work-around to merge /var/lib/ghc-*/package.conf.d
into /var/lib/ghc/package.conf.d. Also remove version number from
/usr/lib/ghc path.
* Fix ghc-ghci existance check logic
* New upstream release (Closes: #622731)
* Mention correct haddock version in package description.
* Install .haddock files under /usr/lib/ghc-doc/haddock, a path valid across
ghc-versions. Also move them to the ghc-doc package.
* In ghc-doc.prerm, do a better job of cleaning
/usr/share/doc/ghc-doc/html/libraries of files generated when creating the
haddock index.
* Make Provides: ghc-ghci dependent on existence of ghci binary
(Closes: #629038)
* Make ghc-haddock depend on the exact version of ghc, just to be on the
safe side. (Closes: #539312)
* Remove Build-Conflict on dash, as a fixed package has been uploaded.
* Do not provide utf8-string or xhtml, as they are only built internally.
* Remove Build-Depend alternative ghc6, and also remove Provides: ghc6, as
this is a lie and I see no techical reason for it.
* New upstream release
* Avoid "hPutChar: resource vanished" message and simplify postinst script
(Closes: #619403)
* Remove package cache in prerm, even on upgrade. This is the right thing to
do if upgrading to a new upstream release, and does not hurt in the other
cases.
* Upload to unstable
* Set a dummy HOME variable upon build. Some build daemons do not set HOME,
but ghc-cabal expects it to be available.
* Build-Conflict with broken version of dash (See #618023)
* Link ghc-doc to haddock via virtual package "haddock-interface-<n>"
* Remove redundand Build-dependency on autoconf
* ghc6-preinst: Remove package.cache file of any ghc6 installation to avoid
dpkg warning at upgrade time.
* Build-conflict with ccache. pbuilder installs ccache by default, which
causes ghc to pick up a wrong path for "gcc".
* Pass -lpthread everywhere. This can possibly be dropped later.
[ Iain Lane ]
* debian/rules: Build with -mminimal-toc on ppc64 — thanks to Colin
Watson for investigation and patch.
[ Joachim Breitner ]
* Disable the use of epoll_create1 and use epoll_create, as the former is
not available on Linux kernels earlier than 2.6.27.
http://hackage.haskell.org/trac/ghc/ticket/5005
[ Joachim Breitner ]
* Remove procps dependency (watcher.hs has been removed since a while)
* Build-depend and depend on libgmp-dev instead of libgmp3-dev.
[ Erik de Castro Lopo ]
* Add patch to fix powerpc compile error (powerpc-compile-616635)
(Closes: #616635)
[ Joachim Breitner ]
* fix FTBFUS by removing left-over binaries
* (Closes: 554174) Segfaults at startup not observable any more
* Tweak debian/copyright
* New upstream release (Should help to close 616494)
[ Joachim Breitner ]
* New upstream release, some patches included.
* Package adopted by the Debian Haskell Group. Thanks to kaol for his
work!
* Rename source and binary packages to ghc, include transitional dummy
package
* Standards-Version 3.9.1 (no changes necessary)
[ Erik de Castro Lopo ]
* Re-enable building of ghci on powerpc. (Closes: 557185)
* New upstream release
* ghci :m now works with module names containing ' (Closes: #580568)
* Put haddock and the shared libraries into their own packages.
(Closes: #571537)
* Only insert packages in gen_contents_index which are both registered
and for which the html directory exists.
* Use ghc-pkg --global in ghc6's trigger to avoid accessing HOME.
(Closes: #578679)
* Fix directories in haddock manpage. (Closes: #578321)
* Apply some patches from Gentoo (thanks, Sergei Trofimovich) hopefully
fixing ia64 build.
* Downgrade ghc6's dependency on perl to Suggests and have ghc6-doc
depend on perl. (Closes: #553332)
* Standards-Version 3.9.0 (no changes necessary)
* ghc6 package now provides ghc6-ghci on architectures where it's
supported (i386 and amd64 only currently).
* Remove debian/watcher.sh. (Closes: #571824)
* Don't ship /var/lib/ghc-6.12.1/package.conf.d/package.cache with ghc6.
(Closes: #574417)
* Patch compiler/rename/RnSource.lhs to make hlist happy.
* Remove some more generated .hc files in debian/rules clean.
* Don't ship broken symlink /usr/lib/ghc-6.12.1/package.conf
* Fix the kfreebsd-* patch.
* Don't die in ghc6's trigger if ghc-pkg check fails. (Closes: #570921)
* Recognise DEB_BUILD_OPTIONS=parallel.
* Add "unset LC_ALL\nexport LC_CTYPE=en_US" to bin/ghc wrapper on ia64.
* Actually append GhcWithInterpreter=YES to mk/build.mk on kfreebsd-*,
don't just echo it to stdout.
* Use dh_haskell_depends from haskell-devscripts to create provides
including the packages ABI.
* Make provided substvar for utf8-string too.
* Run "ghc-pkg check" in ghc6's trigger also.
* Added a stub /usr/share/doc/ghc6-doc/index.html that redirects to
html/index.html.
* Remove /usr/lib/ghc-$(ProjectVersion)/lib/haddock (ie. the internal
haddock binary that ghc6 used at build time) from the ghc6 package.
* Install haddock's files from /usr/share/haddock-$VERSION in ghc6
package, not ghc6-doc.
* Use tabs instead of 8 spaces in debian/rules for ia64 specific flags.
* Include patches to haddock to ghc6's own patches and don't apply them
in debian/rules.
* Update the patch for alpha.
* ghc6-doc depends on ghc6 (>= ${binary:Version}), not haddock.
* Upload to unstable.
* Use some more compiler flags to make ghc6 build on ia64 for real.
* Patch TcForeign.lhs and EventLog.c to make ghc6 build on alpha.
* Included haddock with the ghc6 package. Provides, conflicts and
replaces haddock. Copied over haddock.1 from the haddock package.
* Build with UseLibFFIForAdjustors=YES on alpha.
* Set ghc6 depend on libc6-dev. (Closes: #567809)
* Standards-Version 3.8.4, no changes necessary.
* Patch up the sources some more for kfreebsd. (Closes: #567298)
* Patched utils/haddock/src/Haddock/Interface/Create.hs to use
lookupGlobalName instead of lookupName when GHCI is
unavailable. (Closes: #566331)
* Updated ghc-pkg's man page.
* Added a trigger to run ghc-pkg recache when files are added to
/var/lib/ghc-6.12.1/package.conf.d/.
* Patch rts/posix/OSThreads.c to define _GNU_SOURCE on kfreebsd
too. (Closes: #565818)
* Patch utils/haddock/src/Haddock/Interface/AttachInstances.hs to use
Nothing as mb_info in attachInstances when GHCI is
unavailable. (Closes: #566331)
* Breaks cabal-install (<< 0.8.0).
* Added missing build dependency: libncurses5-dev.
* Use the included haddock to build documentation and drop the
build-depends-indep on haddock.
* New upstream release (Closes: #539789)
* System.Posix.Semaphore fixed. (Closes: #544127)
* Switch to dpkg-source 3.0 (quilt) format
* Standards-Version 3.8.3 (no changes necessary).
* Added a dependency on libbsd-dev (Closes: #549403)
* Patched compiler/utils/Panic.lhs to add a message to internal GHC
error messages, suggesting removing .hi files and pointing to
README.Debian. (Closes: #532443)
* Don't alter Binary.hs and just let Ints to be stored in architecture
native format.
* Don't set XMLDocWays in build.mk since it's not used anymore.
* Remove lpia arch from debian/rules.
* Split debian/rules binary to binary-arch and binary-indep. (Closes: #543972)
* Set SRC_HC_OPTS += -lffi in build.mk and update the patch
system-libffi to use Debian's packaged libffi.
* Update the patch to make the build system use /usr/bin/haddock and not
build its own.
* Patch libraries/unix/System/Posix/Semaphore.hsc
* New upstream release
* Pass UseLibFFIForAdjustors=YES to build options on armel.
* Standards-Version 3.8.2 (no changes necessary).
* Pass GhcUnregisterised=NO to build options on lpia (for Ubuntu's
benefit).
* Patch mkWeakForeignEnv# in rts/PrimOps.cmm to avoid random
segfaults. (re:
https://bugs.launchpad.net/ubuntu/+source/ghc6/+bug/382803)
* Exclude internal boot libraries from generated Provides: libghc6-*
substvars. (Closes: #531318)
* Only call ghc-pkg6 in ghc6-doc's trigger if ghc6 is installed.
(Closes: #530732)
* Fix the test for seeing if ghci would work.
* Build a registerised build on powerpc, but disable ghci.
* New upstream release
* GHCi uses stdio in blocking mode now. (Closes: #512762)
* GHC can cope with missing setitimer(ITIMER_VIRTUAL) support.
(Closes: #509252)
* Uses haskeline instead of editline.
* Only provide ghci and runghc on archs that support it.
(Closes: #320335)
* Patched compiler/utils/Binary.hs to store Ints as Int64s on 32 bit
architectures and warn if it needed to truncate when unserialising.
* Moved package.conf to /var/lib/ghc-$VERSION/.
* Added a /var/lib/ghc-$VERSION/package.conf.d/ dir.
* Set symlinks from /usr/lib/ghc-$VERSION/package.conf* to /var.
* Don't ship generated index files in /usr/share/doc/ghc6-doc/libraries/
and remove them in postrm. (Closes: #501188)
* Remove any leftover index.html files in
/usr/share/doc/ghc6-doc/html/libraries/. (Closes: #461323)
* Bumped to debhelper compat 7 and Standards-Version 3.8.1 (no changes
needed).
* Section: haskell for ghc6 and ghc6-prof.
* Fix up generated Provides for ghc6 and ghc6-doc.
(Closes: #514085, #518400)
* Added man pages for runghc and ghc-pkg. (Closes: #460425, #315763)
* Removed alternative for /usr/bin/ghcprof. (Closes: #527382)
* Build an unregisterised build on powerpc. (Closes: #514946)
* Build-depend on a newer binutils on [arm armel]. Add -mlong-calls to
gcc's flags.
* gen_contents_index reintroduced, this time as a perl script.
* Added /usr/share/ghc6-doc/ghc-$VERSION/desc/ for copies of package
conf info in -doc packages.
* Added a preinst for ghc6, to remove a package.conf file from a removed
but not purged old version.
* Haddock again just a B-D-I, now with (>= 2.4.1-4).
* Patched compiler/utils/Binary.hs to store Ints as Int32s on 64 bit
arches, too.
* Put .haddock files back to ghc6-doc.
* Put haddock in Build-Depends on [i386 amd64 sparc powerpc mips mipsel
s390 kfreebsd-i386].
* Put .haddock files in ghc6, not ghc6-doc.
* Don't call ghc-pkg in a loop in ghc6-doc's trigger, use perl instead.
* Don't remove packages.conf on upgrade from one 6.10.1+dfsg1 version to
another.
* Generate haddock index in ghc6-doc's postinst configure, too.
* chmod +x debian/mk_provided_substvars before calling it.
* Made ghc6-doc's postinst only call haddock on haddock files that are
actually installed.
* Moved xsltproc, docbook-xsl, docbook-xml back as Build-Depends.
* Set build deps related to doc building as Build-Depends-Indep.
* Set BUILD_HADDOCK_DOCS at build time, depending on haddock's
presence. See debian/rules for rationale.
* Replaced /usr/lib/ghc6-doc/gen_contents_index with a symlink to
/bin/true.
* Amended ghc6-doc's "postinst triggered" to perform what g_c_i did.
* Added GhcDebugged=YES and some other flags to build.mk on ia64 and hppa.
* Moved provided-{dev,prof,doc} generation to its own script and catch
errors in it. (Closes: #514085, #514086)
* Further cleanups to debian/rules.
* This time actually change the build on ia64 to be unregisterised.
* Dropped ghc6's dependency on haskell-utils.
* Removed calls to haskell-utils from prerm and postinst.
* Build haddock docs on i386 and amd64. (Closes: #514088)
* Add libffi-dev as a dependency for ghc6. (Closes: #513289)
* Replaced libreadline5-dev with libedit-dev from ghc6's deps.
* Don't add any extra flags to GhcRTSWays build variable.
* Build a registerised build on kfreebsd-i386. (Closes: #513198)
* Build an unregisterised build on ia64.
* Add pkg-config as a build dep.
* Explicitly build an unregisterised version of the compiler on other
arches but i386, amd64, powerpc, ia64. (Closes: #512827)
* Remove --relax altogether from ia64's ld flags.
* Again, use the same build options for all arches and just use what
arch-specific exceptions upstream provided.
* Patched the build system to use haddock from /usr/bin/, not build its
own.
* Disabled building haddock docs for this version.
* Re-enabled some of 6.8.2's build options for non-i386, non-amd64.
* Patch compiler/Makefile's ia64 build options;
s/--ld-option=-Wl,--relax/--ld-option=--relax/.
* New upstream release (Closes: #495126)
* Change the calling conventions for unboxed tuples slightly.
(Closes: #365497)
* Better documentation for swapMVar. (Closes: #405717)
* Don't change code in error messages. (Closes: #499137)
* Improve error reporting for 'deriving' (Closes: #499216)
* Better error message when -XRankNTypes is missing. (Closes: #499217)
* ghc-pkg respects --global with 'field' option. (Closes: #510499)
* Repackaged to remove a copy of GNU MP library.
* Enable building the stage2 compiler on all architectures.
* Added information about libffi (which is included with GHC) to
debian/copyright.
* Manage changes to the source with quilt.
* Link against the system's libffi and add libffi-dev as a build
dependency.
* Build-dep on haddock >= 2.4.1-1.
* Build-dep on libedit-dev, removed build-dep on libreadline-dev.
* Patched gen_contents_index: fixed the case when not run inplace; trac
#2764
* Patched libraries/base/Data/Data.hs: use Prelude.(,,) for
tuple3DataType; trac #2750
* Added a trigger to ghc6-doc for /usr/share/doc/ghc6-doc/libraries to
run gen_contents_index. (Closes: #506568)
* Repackaged the upstream tarball to remove a copy of GNU MP library
with GFDLed docs w/ invariant sections. (Closes: #511756)
* Build-Depend on hurd as an alternative to procps. (Closes: #481343)
* Added AC_SYS_LARGEFILE to libraries/unix/configure.ac and ran
autoreconf. (Closes: #500407)
* New maintainer.
* Made the perl script driver/split/ghc-split not use the obsolete $*
var (Closes: #489157)
* Copied libraries/unix/System/Posix/Resource.hsc and
libraries/base/include/HsBase.h from 6.8.3 to fix issues with
setResourceLimit. (Closes: #491909)
* Don't build template-haskell if we're not building GHCi.
The package is largely useless without GHCi, and some of the buildds
were having trouble building template-haskell. We'll need to fix
this some other way if GHCi is to be available on every arch, though...
* Small wibbles to debian/watcher.sh.
* Add a build-dep on procps (debian/watcher.sh runs ps).
* Every 10 minutes, print any "ps ux" lines that mention gcc or ghc.
According to folks on IRC, this is standard practice. It means that
we don't have to worry about security buildds having different
timeouts to the normal builders.
* Apply upstream patch:
FIX #2073: Don't add empty lines to GHCI's history
Ian Lynagh <igloo@earth.li>**20080224143256
Closes: #461170.
* Apply upstream patch:
Tweak the splitter
Ian Lynagh <igloo@earth.li>**20080116195612
We were generating a label ".LnLC7", which the splitter was confusing
with a literal constant (LC). The end result was the assembler tripping
up on ".Ln.text".
Closes: #466262.
* Make an hpc symlink in /usr/bin. Closes: #461146.
* Add a dep and build-dep on gcc >= 4:4.2 as we need the
-fno-toplevel-reorder flag. Closes: #461332.
* Put the right upstream source URL in debian/copyright.
Closes: #465058.
* Strengthen haddock dep and build-dep to 0.8-2 as we need the
version that makes compatible files on 32 and 64 bit arches.
Closes: #433251.
* Add a build-dep on hscolour. We now use it when making the
documentation, so the haddock docs link to syntax highlighted
source.
* Remove the libsrc package. ghc6-doc now includes the sources anyway.
Closes: #432706.
* gen_contents_index is now provided by the upstream install, so we
no longer ship our own copy. However, we do have to move it into
/usr/lib rather than /usr/share/doc.
* Remove the stat2resid slave of the GHC symlink, as stat2resid
no longer exists. Closes: #432715.
* We set bindir, docdir, htmldir, dvidir, pdfdir and psdir in
mk/build.mk rather than with configure flags, as the way configure
currently works means they can't be set relative to other variables
if set with configure flags.
* We no longer pass --datadir to configure; we used to use this to
get the docs in the right place.
* When doing the install step, we now use DESTDIR rather than prefix
to override where the files go.
* We now need to set HADDOCK_DOCS=YES in mk/build.mk in order to get the
haddock docks built.
* Removed "html/" from the documentation path in ghc6-doc.postinst,
ghc6-doc.prerm and ghc6-doc.doc-base.users-guide.
* Use "[ ! -f mk/config.mk ] ||" to guard "make distclean" rather than
ignoring all errors from it.
* Use ${binary:Version} rather than ${Source-Version} to specify the
dependency of ghc6-prof on ghc6.
* All the libraries now install LICENSE files, so we remove them after
the install step.
* Follow change in man path from /usr/man to /usr/share/man.
* lintian thinks that Cabal's Distribution/License.hi is an extra
licence file, so add a lintian override.
* ghc6-doc now depends on haddock as it needs to run
gen_contents_index in postinst. Closes: #423561.
* ghc6-doc and ghc6-libsrc depend on ${shlibs:Depends}, ${misc:Depends}.
* New upstream version.
* Policy 3.7.2 compliant.
* ghc6-doc provides libghc6-PACKAGE-doc for the various packages it
has docs for.
* Tweaked debian/rules to handle changes.
* Remove build-dep on xutils; we no longer use lndir (which has now
moved to xutils-dev anyway!)
* Remove build-dep on cpio; we now let dh_install do all the moving
rather than doing it ourselves with cpio.
* Remove build-dep on time; no longer used.
* Remove build-deps on libx11-dev, libsm-dev, libice-dev, libxmu-dev,
libxi-dev; the X11 stuff is now in separate Cabal packages.
* Manpages generation script removed s it is now in upstream.
* Pass --datadir to ./configure rather than "make install-docs"
so the haddock fields in package.conf get set correctly.
Closes: #417325.
* In mk/config.mk.in, don't put ghc-6.6.1/ on the end of datadir.
* Use $(INSTALL) rather than hardcoding /usr/bin/install everywhere.
* Don't generate or install library HTML doc contents and index.
* Do install libraries-footer.txt, libraries-header.txt and the various
prologue.txt's.
* Install a gen_contents_index script for generating the haddock
contents and index.
* Add -X.haddock to the dh_compress call to make sure the .haddock
files aren't getting compressed.
* Give ghc6-doc a postinst to generate the haddock contents and index
when it is installed.
* Give ghc6-doc a prerm to clean up the above.
* Revert these earlier changes as we now have haddock 0.8:
* Remove the --source-module argument to haddock in mk/package.mk
as haddock 0.7 doesn't support it.
* Re-add the -optP-P when generating .raw-hs files as haddock 0.7
doesn't cope with line numbers in the files.
* Add arm to the list of arches that have ghc6.
* Add arm to the arches in compiler/cmm/PprC.hs for which
loads and stores to be printed in a way that works if they are not
aligned as the arch wishes.
* For arm's odd floating point numbers:
* Add FPTOOLS_FLOAT_WORD_ORDER_BIGENDIAN test to aclocal.m4
* Call FPTOOLS_FLOAT_WORD_ORDER_BIGENDIAN after AC_C_BIGENDIAN
in configure.ac.
* Extra section for the FPTOOLS_FLOAT_WORD_ORDER_BIGENDIAN test in
configure.
* Add "#undef FLOAT_WORDS_BIGENDIAN" to mk/config.h.in.
* Add FLOAT_WORDS_BIGENDIAN cases to rts/StgPrimFloat.c.
* Apply the following upstream patch, to fix potential problems
compiling ghc6 on amd64 (and possibly others):
Fri Oct 20 16:39:25 BST 2006 Simon Marlow <simonmar@microsoft.com>
* In hashExpr, use Word32 rather than relying on wrapping behaviour of Int
Fixes #952, as it turns out.
When compiling via C, we are at the mercy of C's undefined behaviour
with respect to overflow of signed integer operations, and this was
biting us here.
Perhaps we should always add the -fwrapv flag to gcc, but since
Haskell doesn't define overflow on Int either, it seemed the right
thing to do to fix this code anyway.
* Add mips and mipsel to the list of arches that have ghc6.
* Add mips and mipsel to the arches in compiler/cmm/PprC.hs for which
loads and stores to be printed in a way that works if they are not
aligned as the arch wishes.
* Removed the -static flag for mips from compiler/main/DynFlags.hs.
* New upstream version.
* Drop ghc6-hopengl package as HOpenGL is now split off from the
GHC core.
* Removed:
* ghc6-hopengl.README.debian
* ghc6-hopengl.examples
* ghc6-hopengl.postinst.in
* ghc6-hopengl.prerm.in
* Removed opengl examples.
* debian/scripts.mk no longer makes:
* debian/ghc6-hopengl.postinst
* debian/ghc6-hopengl.prerm
* Removed all opengl-related commands from debian/rules.
* Drop build-deps on
xlibmesa-gl-dev, libglu1-xorg-dev | libglu-dev, libglut3-dev
now we don't build the OpenGL libs.
* Added note to manpage that ghci is not yet available on all arches.
* Add ppc64 to the list of arches that have ghc6. Closes: #375623.
* Update locations of README and ANNOUNCE in ghc6.docs.
* Don't include ghc/mk/version.mk in debian/scripts.mk as it no longer
exists. Instead, get ProjectVersion with some shell magic.
* Don't include ghc/mk/version.mk in debian/scripts.mk as it no longer
exists. Instead, pass $(ProjectVersion) in from debian/rules.
* Remove mk/build.mk before we start filling it so the commands can be
more symmetric.
* Don't build the threaded RTS except on x86/amd64.
* Don't try to link ghc6 with the threaded RTS except on x86/amd64.
* Remove debian/test-build before creating it, so we can restart
builds part-way through without them falling over.
* Various paths in debian/rules lose their ghc/ prefix.
* Clean up debian/test-build/ after doing the test.
* hslibs no longer exists, so remove everything relating to it in
debian/rules.
* Remove register declarations from rts/StgCRun.c that break the
unregisterised build on alpha.
* Add to compiler/cmm/PprC.hs an option for loads and stores to be
printed in a way that works if they are not aligned as the arch
wishes. Enable this option for alpha.
* Remove the --source-module argument to haddock in mk/package.mk
as haddock 0.7 doesn't support it.
* Re-add the -optP-P when generating .raw-hs files as haddock 0.7
doesn't cope with line numbers in the files.
* Have ghc-pkg not create ~/.ghc when it doesn't need to, and return
the empty package file when trying to read a package file that
doesn't exist. Closes: #375166, #375188.
* New upstream release. Closes: #369947.
* Add --nonet to XSLTPROC_OPTS.
* Add GhcWithInterpreter=NO on arches we don't build registerised
due to http://hackage.haskell.org/trac/ghc/ticket/631
* Non-maintainer upload.
* Add docbook-xml to build-dep (Closes: #356015).
* Due to problems with xmltex's uninstallability and its long-standing
bugs with lists rendering some docs unbuildable:
* Set XMLDocWays to "html" in debian/rules, rather than "html dvi ps".
* Remove build-deps on jade, docbook-utils, xmltex, docbook-xml.
* Change ghc6-hopengl's dep on "libc6-dev" to
"libc6-dev | libc6.1-dev | libc-dev"
* Fix building cabal packages that build executables with HS-Source-Dir.
Closes: #337909.
* Use "tail -n +2" rather than the deprecated "tail +2" in debian/rules.
Closes: #339606.
* On the list of arches we build registerised:
* Add amd64. Closes: #354872.
* Remove sparc (it has bitrotted).
List now reads "i386 amd64".
* Remove ghc6.README.Debian (appeal for registerised ports). The problems
when they bitrot outweigh the advantages IMO, at least until we have a
testsuite we can run when building.
* Apply fix for ghci failing to start ("Unable to mmap( MAP_FIXED ) for
Jump Islands") from Ryan Lortie. Closes: #343428.
* Delete ".SECONDARY:" from mk/suffix.mk as make >3.80 was taking
forever due to bug #346248. Closes: #348633.
* Changed xlibs-dev build-dep to:
libx11-dev, libsm-dev, libice-dev, libxmu-dev, libxi-dev.
Closes: #346752.
* Add "GhcWithNativeCodeGen=NO" to mk/build.mk when we are building
unregisterised.
* New upstream release. Closes: #329322.
* Relax the libgmp3-dev build-dep to make backporting easier.
* Don't force using gcc-3.3 to build; drop gcc-3.3 build-dep.
* Apply patch to add support for kfreebsdgnu to configure(.ac) and the
mangler. Closes: #332977.
* Add kfreebsd-i386 to ghc6_arches.
* Add missing deps on libx11-dev, libsm-dev, libice-dev, libxmu-dev,
libxi-dev, libc6-dev to ghc6-hopengl. Closes: #317069.
* Update (build-)dep xlibmesa-glu-dev to libglu1-xorg-dev | libglu-dev.
* Standards version 3.6.1 -> 3.6.2.
* Upstream fixed a panic to give a sensible error message.
Closes: #319294.
* Non-maintainer upload.
* C++ ABI transition: Relink against new libgmp3c2 package.
This will require a manual bootstrapping on each arch since ghc6
build depends on itself and the new and old version of gmp3
conflict with each other. (Closes: #319222)
- Versioned build dependency on libgmp3-dev (>= 4.1.4-6)
* Build using gcc-3.3 since this version doesn't build/work using 4.0
and a new upstream version that fixes it isn't available yet.
See instructions in debian/bootstrap-extracted on how to build this.
* Use &(arr->payload) instead of BYTE_ARR_CTS(arr) in Adjustor.c on
ia64 as we no longer have the definition of BYTE_ARR_CTS to hand.
* Use ASSIGN_DBL and PK_DBL in generated C code from CMM code to avoid
unaligned access SIGBUSes on some arches. Closes: #309025.
* Use the stage2 ghc-inplace when checking "hello world" works.
* Build-dep on xmltex, docbook-xsl and docbook-xml.
* Set XMLDocWays rather than SGMLDocWays. Closes: #309016.
* Disable dvi and ps docs as they don't build.
* Remove ps and dvi sections from debian/ghc6-doc.doc-base.users-guide.
* Removed powerpc from the list of arches to do a registerised build with
object splitting as there are bootstrapping issues from earlier GHCs.
* Change a stray "6.4" to $(ProjectVersion) in debian/rules
* Fix "SOURCE:" lines in {pre,post}{inst,rm}.in files.
* Don't create user package.conf if not modifying.
* Change "#ifdef darwin_REGS" to "#if darwin_REGS" in
ghc/includes/MachRegs.h to fix build on powerpc (from upstream CVS).
* Add powerpc to the list of arches to do a registerised build with
object splitting.
* Fix checkFEDArgs in ghc/compiler/typecheck/TcForeign.lhs so it compiles
again.
* Remove OpenGL and GLUT packages from package.conf.shipped.
Have the postinst/prerm add/remove them.
* ghc6-hopengl provides libghc6-{opengl,glut}-{dev,prof}
* Other exposed libraries are provided as libghc6-foo-dev by ghc6 and
libghc6-foo-prof by ghc6-prof.
* New upstream release; just for experimental for now.
* Remove xmlise-flags.sgml as upstream now ships flags.xml.
* Rejig how files are put in the right packages to follow upstream changes.
* Separate build target into build and install.
* No longer allow ghc4 or ghc5 to be used for building as they are no longer
supported.
* Handle /usr/bin/ghc6 et al. and the alternatives without the 6 suffix
better.
* Add a /usr/bin/runhaskell alternative, priority 8600000600.
* Backport runghc from CVS.
* Build-dep prefers libreadline5-dev rather than libreadline4-dev.
* Compiled with libreadline5-dev. Closes: #291126.
* Remove package.conf{,.old} when changing to a different upstream
version as well as when purging. Closes: #277606.
* Add build-depends on autotools-dev.
* Add "Modes of operation" flags to flag reference in the users guide,
and hence to the automatically generated manpage.
* Create /usr/bin/ghc-pkg6.
* Revert changes to support mips/mipsel. They need changes in
glibc/gcc to work, supported by different changes in ghc.
* Support the spirit of noopt and nostrip values of DEB_BUILD_OPTIONS
* Make symlinks for /usr/bin/{ghc,ghc-pkg,ghci}-$version into
/usr/lib/ghc-$version/bin/.
* Add manpage symlinks for {ghc,ghci}-$version.
* Bump Standards-Version to 3.6.1.
* Remove mention of threaded packages from the manpage as they no
longer exist.
* Add grep-dctrl to build-dep list and use grep-status to build a
depends line for the set of installed packages providing
libreadline-dev (should have cardinality 1).
Replaced libreadline4-dev dependency with ${readline}.
Switch order of libreadline4-dev and libreadline-dev in build-deps
so the concrete package is first.
* Change how configure/config.sub/config.guess are handled in line
with /usr/share/doc/autotools-dev/README.Debian.gz.
* Added varfile /usr/lib/haskell-utils/ghc6_vars.
* Remove docs/building/building.out and
ghc6-6.2.2/docs/docbook-cheat-sheet/docbook-cheat-sheet.out when
cleaning.
* Join build-deps onto a single line.
* Eliminate {,} bashisms in debian/rules.
* Build-dep on xlibs-dev for building the X11 package.
* Additionally build the RTS the following ways:
debug_p thr_debug thr_debug_p
(which, together with those done by default, gives us all ways).
* Install initial /usr/lib/ghc-$version/package.conf as
/usr/lib/ghc-$version/package.conf.shipped and copy it to
/usr/lib/ghc-$version/package.conf in postinst if it doesn't exist.
This means other library information isn't lost across upgrades.
* Remove /usr/lib/ghc-$version/package.conf{,.old} in postrm purge.
* Remove IA64 from the unregisterised arches list to try to fix the
unaligned accesses.
* Add -mlong-calls flag for mips* in both DriverFlags and SRC_CC_OPTS
in mk/config.mk.in. Fixes problem where a step in gdb on a call to
mp_set_memory_functions lands us in the middle of Show.hc.
* Up gcc build-dep to "gcc-3.3 (>= 1:3.3.4)" to fix
"charToUtf8 1884139800" panic when compiling
'module Foo where foo 5 = 6'.
* Change the mangler to allow a tab before .section on sparc.
Fixes a problem which shows up as symbols not being made global
so not being defined when compiling with gcc >= 3.something.
* New upstream (stable branch) release.
* Don't use -static on any arches (remove per-arch "-static"s in
ghc/compiler/main/DriverFlags.hs).
* Remove threaded packages (functionality now handled by main packages).
* We now depend and build-depend on "gcc-3.3 (>= 1:3.3.3-2)" which should
give correct code on all arches.
* More HOpenGL example tweaking.
* Apply fix so OpenGL docs build when the compiling compiler doesn't
support OpenGL.
* Tweaked HOpenGL example, including suggestions from Sven Panne (HOpenGL
author).
* New upstream version.
* ghc/docs/users_guide/flags.sgml is now xmlised by a script rather
than by hand.
* Added an HOpenGL example.
* Removed build dependency on happy (it has been pre-run on the tarball).
* Build dependency on haddock is now >= 0.6 (we need some of its new
options).
* For the normal ghc6 deb pass GhcLibsWithOpenGL=NO GhcLibsWithGLUT=NO
rather than GhcLibsWithHOpenGL=NO
* Add config.status and libraries.txt to CLEAN_FILES and a rule to
remove the HTML directory for the extraclean target in
libraries/Makefile
* Change #ifdef IN_STG_CODE to #if IN_STG_CODE in the alpha section of
ghc/includes/TailCalls.h
* Add __muldi3 to RTS_LIBGCC_SYMBOLS in ghc/rts/Linker.c (ghci was
complaining about it being unknown on sparc).
* Provide a /usr/bin/haskell-compiler alternative.
* Depend on haskell-utils
* Register /usr/bin/ghc6 and ghc6 with haskell-utils
* Register /usr/bin/ghc6-threaded and ghc6-threaded with haskell-utils
* Tidy up postinst/prerm.
* Add ghc6-threaded as an alternative for /usr/bin/ghc (priority 590,
i.e. between ghc5 and ghc6).
* Move the TailCalls.h include in ghc/includes/Stg.h below the config.h
include. Fixes the undefined reference warnings in -8. Closes: #211430.
* If we are not on an i386, sparc or ia64 machine then do an
unregisterised build.
* Unregisterised builds should be possible for any architecture, so
set the architecture to any.
* Add s390, m68k, mips, hppa, arm and powerpc Linux to configure{,.in}
* Write a README.Debian about the registerised/unregisterised deal.
* Change a couple of machine/foo.h includes on alpha to asm/foo.h
* Add the x86-64 hack from upstream to MBlock.h, generalising it to all
8-bit arches.
* Add -optc-mbig-switch to SRC_HC_OPTS in libraries/OpenGL/Makefile
on hppa (fixes an assembler failure for at least
Graphics/Rendering/OpenGL/GL/QueryUtils.p_o).
* Tweak ghc/includes/TailCalls.h so it only steals a register for STG
code.
* Move the include of TailCalls.h in ghc/includes/Stg.h up so it is
before all the procedure definitions.
* Change an IF_OS_darwin to an IF_ARCH_powerpc in
ghc/compiler/nativeGen/MachCode.lhs
* Add ia64 to Architecture field of the packages.
* Apply various IA64 fixes from upstream CVS.
* We need docbook-utils even for a binary-only build as the build fails
if it can't build the documentation.
* Add jade to the build-deps.
* Remove unnecessary flex build-dep and comment it out of
configure{,.in}.
* Resurrect the "SGMLDocWays := html dvi ps" line in build.mk that
got lost.
* Do a test compilation+run after building to make sure things aren't
completely screwed up.
* make distclean rather than maintainer-clean
* Put sparc-unknown-linux entry in configure and configure.in
* Add sparc Linux support to ghc/driver/mangler/ghc-asm.lprl
* Build depend on gcc-2.95 [sparc]
* Change ghc6's dependency on gcc to ${gcc}.
This is set to either gcc-2.95 (if we are on a sparc) or gcc
(otherwise) in debian/ghc6.substvars
* Removed quotes around ${WithGhc-ghc} in configure and aclocal.m4 so
arguments can be given (e.g. to point at a certain gcc) with the
flag
* Pass flags giving the location of ghc and, in the case of sparc,
gcc to ./configure
* Added sparc to Architecture field of the packages
* touch configure early in the build and clean targets to stop the
build system trying to regenerate it with autoconf.
* Ask for the correct name to be removed by update-alternatives
* Have manpage variables substituted
* Fix problem with files not getting installed into ghc6-libsrc package
* Put ghc/{README,ANNOUNCE} in /usr/share/doc/ghc6
* New upstream version
* Now policy 3.6.0 compliant
* Fixed manpage alternatives to create a manpage for ghc, not ghc6.
* Added manpages for ghci, ghci6 and the -threaded variants of
these and ghc. Manpage now also mentions ghci.
* Reinstate the -libsrc package.
* Remove redundant build dependencies on autoconf and autotools-dev
* Comment out fix for compiling GL stuff with older GHCs as it really
uses the in-place compiler
* Have all packages (except ghc6-doc due to issues with an empty depends
line) depend on ${shlibs:Depends}, ${misc:Depends}
* Create a /usr/bin/ghc6 and /usr/bin/ghci6
* Also build a threaded-rts binary and make various related -threaded files
and symlinks. Put in a separate package (ghc6-threaded-rts) due to size.
* Fix GET_PROC_ADDRESS properly so opengl libraries can be used (from
upstream CVS)
* GLUT libraries are also installed in the hopengl deb
* Update build-depends and depends to libglut3-dev, xlibmesa-gl-dev
and xlibmesa-glu-dev from glutg3-dev and xlibmesa-dev.
* Require at least version 0.4.0 of haddock.
* Fix doc-base control file links.
* Policy 3.5.10 compliant.
* Create the hp2ps-ghc6.1 manpage correctly. Closes: #199786.
* Change glActiveTexture to glActiveTextureARB. ghci now loads with
the OpenGL and GLUT packages.
* Update build-depend on debhelper to >= 4.
* Correct "GHC5" to "GHC6" in the doc-base title.
* New upstream version
* New maintainer, with the old maintainer's blessing
* changed package name to ghc6
* Building process simplified as GHC's build system now handles multiple
stages itself
* Patches from upstream:
+ Removed comments from cpp functions in a couple of source files to
counteract changes in 3.3 meaning they didn't compile
+ Evil mangler tweaked to allow extra whitespace gcc 3.3 puts in
and for changes to generated assembly for _module_registered
+ Rename "--list-packages-local" to "--list-local-packages"
* Added missing quotes to
libraries/OpenGL/Graphics/Rendering/OpenGL/GL/Extensions.hs
so it compiles
* Removed fptags as it doesn't have a clear licence; should probably
be in its own package anyway as I imagine it is equally useful with
any Haskell compiler
* Changed libraries/OpenGL/specs/enumerant to use this compiler rather
than the installed one; in particular this means that the profiling
libraries don't need to be installed
* CVS Version from 2002-12-05 (tag: ghc-5-04-2)
+ fix for inet_ntoa linking error (upstream)
(closes: Bug#164573)
+ export TimeLocale constructor (upstream)
(closes: Bug#167647)
* implemented proper auto* handling (config.*, autoconf)
* added autotools-dev to build-deps, bumped needed autoconf version to
>= 2.52
* removed cpio from Build-Deps (no longer needed)
* added man page for ghc
(closes: Bug#110037)
* added links for user_guide/user-guide.html and hslibs/book-hslibs.html
(closes: Bug#153795)
* necessary adjustments for debian_libraries_hopengl.dpatch
* libdir adjustments (always in sync with Debian package version now)
* added notes in ghc5-hopengl.README.Debian
* removed cygnus-stylesheets from Build-Deps-Indep
(closes: Bug#170112)
* moved ghcprof icons to /usr/share/ tree
(debian_ghcprof_datadir.dpatch)
* edited debian/copyright to make lintian shut up ("Upstream Author(s)")
* backport of "*_stub.o link failure" fix from ghc-5-04-branch
(debian_compManager_findstubs.dpatch)
(closes: Bug#171518)
* CVS Version from 2002-07-15 (tag: ghc-5-04)
(closes: Bug#153020)
+ spurious ghci dynamic linkage error solved by upstream update
(closes: Bug#151126)
* added Depends: libreadline4-dev (though I don't like it
and it just covers a technical issue), because too many people
get bitten by this :(
(closes: Bug#134950)
* build process revamped
+ much easier to understand/maintain (KISS)
(it's a build script, no need to be too fancy)
+ support for full bootstrap builds (needed for buildds)
+ moved source to upstream-ghc/
+ adjusted patches; must contain `upstream-ghc/' in
the path from now on (new source layout)
+ uses shadow trees (lndir)
+ added "-e noidref" to docbook2* tool calls, so that they
don't barf on missing ids (there are 2 currently)
+ use upstream doc install mechanism now that it's usable
for Debian (slight changes in the dir layout, though)
+ removed support for `debian/rules cvs-update'
+ removed ghc5-libsrc package, now that decent library
docs start to pop up
+ removed debian_build.dpatch (now debian/build.mk-stage*)
* renamed ghc4_ghc_includes_gmp.dpatch debian_ghc_includes_gmp.dpatch
because it's Debian specific
* added Build-Depends: haddock
* adjusted *.doc-base.* again to reflect new doc layout
(only ghc5-doc.doc-base.users-guide left)
* _really_ changed CVS/Root to :pserver:anoncvs@cvs.haskell.org:/cvs
this time (no idea what happened to the last changes)
* updated CVS checkout instructions in copyright
* updated upstream authors in copyright
* updated README.Debian to reflect changes
* fixed missing "sec-concurrent-haskell" SGML id
* added hasktags program to update-alternatives
* new package ghc5-hopengl (finally!)
(closes: Bug#116984)
* added Build-Depends: xlibmesa-dev, glutg3-dev (for HOpenGL)
* CVS Version from 2002-02-14 (tag: ghc-5-02-2)
(closes: Bug#133595)
* buggy Makefile which prevented building with make-3.77 fixed
by update
(closes: Bug#117883)
* The "You Asked For It" release =)
* _NOT_ the "I Solved All Problems" release
* changed CVS/Root to :pserver:anoncvs@cvs.haskell.org:/cvs
* updated copyright
* removed lib6-dev from Depends: (no longer necessary)
* docbook build-deps now: docbook-utils | cygnus-stylesheets
(closes: Bug#123180)
* added time to Depends: (lost during backporting potato stuff)
* added build-deps: xutils (for lndir), time (used by build process)
* added build-deps: devscripts (for debchange, thanks to
Peter Gammie)
* readline build-deps now: libreadline-dev | libreadline4-dev
* priority: extra -> optional (as requested from the ftpmasters)
* for the -prof package, section: libs -> devel (as requested from
the ftpmasters)
* spelling correction in control file (cpu -> CPU)
(closes: Bug#124668)
* removed getSocketOption patch
(ghc5_getSocketOption.dpatch)
* added index.html symlinks for html docs
* adjusted *.doc-base.*
* removed ghc5-doc.doc-base.hslibs (accessible via
the user's guide)
* bumped Standards-Version to 3.5.6.0
* CVS Version from 2001-10-15 (tag: ghc-5-02-branch)
* adjusted debian_glafp-utils_docbook.dpatch
* typo in docs fixed
(closes: Bug#110038)
* added hotfix for getSocketOption b0rkage (already fixed in CVS)
(ghc5_getSocketOption.dpatch)
* CVS Version from 2001-07-16 (tag: ghc-5-00-2)
(closes: Bug#106756)
* upstream fix for .ghci problem
(closes: Bug#94739)
* removed patches which went into upstream
(ghc5_glitches.dpatch)
* added dependency on libc6-dev as a workaround
(ghci needs the libm.so link)
* adjusted ghc5_glitches.dpatch slightly
* CVS Version from 2001-04-11 (tag: ghc-5-00)
* changed package name to ghc5
* some more reorganizations to prepare dual ghc4/ghc5 installation
* added some missing build-deps
* now using libgmp3(-dev)
* removed some obsolete patches
(ghc4_hclose_bug.dpatch,ghc4_signal_bug.dpatch)
* added some other patches
(ghc5_glitches.dpatch)
* updated build.mk-patch once more
* DH_COMPAT=2
* added more docs (ffi-art)
* added support for CVS happy's
* first upload as proper maintainer
* changed email address to @debian.org
* unofficial version
* backported signal-bug patch from CVS trunk
(ghc4_signal_bug.dpatch)
* unofficial version
* backported hClose-bug patch from CVS trunk
(ghc4_hclose_bug.dpatch)
* !!NOTE!!
This version replaced aaronv's upload of 4.04.19990916-3 (thanks
to wli@debian.org for the sponsor upload). All versions in
between were never officially available on the debian servers.
* CVS Version from 2000-09-02 (4.08.1)
* official 4.08.1 version
* uncommented `cvs upd' commands (accidently commented)
* changed `cvs-update' target again (first step of upcoming
"experimental" support)
* updated build.mk-patch
* set.{ps,dvi} accidently were not copied into the package,
(fixed now)
* CVS Version from 2000-06-29 (4.08.0)
* this is a snap of the official 4.08 release candidate
* need (happy >= 1.7) from now on
* unset DOCBOOK_PREFIX in order to use installed db tools
* an installed ghc4-prof pkg. must have the same version as the
corresponding ghc4 pkg.
* added option `-k' to make when cleaning, so it won't stumble over
errors
* ghc now understands option '--numeric-version', which will be
supported in the upcoming unstable version anyway...
(ghc4_numeric_version.dpatch)
* sync'ed base documents in doc-base entries with reality
* rearranged `docs',`build' targets
* rearranged documentation to sync with upstream
* CVS Version from 2000-05-26 (4.07)
* update to ghc-4-07-branch
* disabled debian_glafp-utils_docbook.dpatch, because it's not
necessary on this branch
* !! UNRELEASED due to building problem !!
* CVS Version from 2000-05-23 (4.07)
* removed ghc4_docbook.dpatch (incorporated by upstream, finally :-))
* removed ghc4_hslibs_tools_clean.dpatch (incorporated by upstream)
* removed ghc4_hslibs_util_clean.dpatch (incorporated by upstream)
* removed building of docbook (Debian does have a working docbook
environment)
(debian_glafp-utils_docbook.dpatch)
* added fptags to distribution (in a really q'n'd way)
* added debian/rules options (UN)PATCHOPTS, default is now
`PATCHOPTS = --ignore-whitespace --forward'
* added notes about debian/rules options to README.Debian
* adjusted build depends
* added more hints in README.debian
* !! UNRELEASED due to building problem !!
* CVS Version from 2000-05-04 (4.07)
* hslibs/utils/Readline_stub.p_o not cleaned
(ghc4_hslibs_util_clean.dpatch)
* added DrIFT, DtdToHaskell, Xtract as update-alternatives
* CVS Version from 2000-04-29 (4.07)
* removed ghc4_hslibs_util_clean.dpatch (solved by upstream)
* Main.hi not cleaned in hslibs/tools/* dirs
(ghc4_hslibs_tools_clean.dpatch)
* changed clean targets from `realclean' to `maintainer-clean'
(due to upstream change)
* !! UNRELEASED due to stage1 problems !!
* CVS Version from 2000-04-11 (4.07)
* removed ghc4_configure_readline.dpatch (incorporated by upstream)
* adjusted Build-depends (this version can only be build with GHC 4.06)
* `debian/rules patch' now doesn't try to apply patches, if they're
already applied
* added option `StripLibraries = YES'
* !! UNRELEASED due to stage2 problems !!
* CVS Version from 2000-02-02 (4.07)
* added target `cvs-update' (see README.Debian)
* replaced sgml-tools by cygnus-stylesheets in Build-Depends
* removed ghc4_hslibs_cstring.dpatch (incorporated by upstream)
* removed ghc4_ghc_mk.dpatch (incorporated by upstream)
* removed debian_sgml.dpatch (no longer necessary)
* use autoreconf instead of autoconf (in debian/rules.local)
* html docs back in action
* renamed doc-base ID ghc4-installation to ghc4-building
* YARP [Just Another Readline Patch]: doesn't find termcap
(ghc4_configure_readline.dpatch)
* Readline_stub--stamp was not removed on clean
(ghc4_hslibs_util_clean.dpatch)
* removed hardcoded DTD paths in docs/fptools-both.dsl
(ghc4_docbook.dpatch)
* added `docs' target
* `debian/rules clean' again checks for root priv. (cf. packaging#3.2.1)
* added build option `-split-objs' (produces smaller executables)
* added build option `GhcLibHcOpts = -O' (won't compile otherwise, because
of a compiler bug in 4.06)
* CVS version from 2000-01-09 (4.06)
* debian/rules clean doesn't check for root priv. any more
* removed emacs local variable (add-log-mailing-address) in changelog.
(closes: Bug#54515)
* references to `haskell-doc' in control and debian/doc-base.
(closes: Bug#52729)
* Arch back to `i386' only, although sparc should be possible, too (with
minor changes, really). (closes: Bug#52794)
* bumped Standards-Version to 3.1.0
* introduced build dependencies
* removed ghc4_clean.dpatch (incorporated by upstream)
* removed ghc4_makefiles.dpatch (incorporated by upstream)
* removed ghc4_mk.dpatch (incorporated by upstream)
* removed ghc4_readline.dpatch (incorporated by upstream)
* removed ghc4_stat2resid.dpatch (incorporated by upstream)
* autoconf changes to rules, rules.local
* corrected bug in ghc/mk/paths.mk (wrt. variable referencing)
(ghc4_ghc_mk.dpatch)
* corrected bug in hslibs/CString.lhs (ghc4_hslibs_cstring.dpatch)
* added patch for sgmltools, removed {html,txt} because of problems
* CVS version from 1999-12-06 (4.05)
* removed distrib/Makefile-bin.in patch (solved by upstream)
* removed ghc/compiler/Makefile patch (solved by upstream)
* adjusted mk/target.mk patch
* moved $(libdir)/ghc/$(version) -> $(libdir)/ghc-$(version) (upstream-like)
* dropped users-guide.info (wasn't complete, anyway)
* added postscript and plain text docs
* heap size back to 120M :-(
* changed the build process heavily
* fixed `missing symbol in libHSutil*.a' (formerly libHSmisc*.a) bug
* some cleanup patches
* updated URLs to the Haskell report in control and debian/doc-base
* added warning message about fragile `clean' target
* cut down memory req's for building GHC (100M heap should be ok)
* Arch: field back to ``i386''
* First revision to be uploaded to Debian
* Changed maintainer fields where necessary
* Changed binary packages to Architecture: any
* fixed typo in copyright
* bumped Standards-Version to 3.0.1
* doc-base support
* moved (un)patch rules to debian/rules.local
* added haskell-doc, ghc4-prof to Suggests
* new version numbering scheme
* improved build process (targets stage1,clean)
* added libgmp2-dev to Depends
* CVS version from 1999/09/16
* fixed manpage packaging bug in ghc4-prof
* Standards-Version 3.0.0.0
* removed undocumented manpages (lintian complains)
* moved usr/info -> usr/share/info according to FHS
* added 'perl | perl5' to Depends
* removed patch for happy version check (now in upstream)
* GHC is working again
* CVS version from 1999/08/03
* local version, GHC doesn't work
* removed hstags from alternatives because it's non-functional
* removed mkdependHS from alternatives because directly running it
is deprecated now (now 'ghc -M')
* re-added info file
* removed obsolete HAVE_READLINE variable in mk/build.mk
* removed patch for {hscpp,mkdependHS} "absolute path" fix (now in upstream)
* moved to patch system (taken from glibc package)
* CVS version from 1999/07/31
* got rid of absolute cpp paths in {hscpp,mkdependHS}
* removed info files due to problem with sgmltools
* first glibc-2.1.1 version
* moved profiling libs into separate package
* added hscpp as alternative (useful for hugs)
* CVS version from 1999/07/10
* compiled with use of happy-1.6 parser generator
* fixed some "unclean" directories
* added info entry
* added dwww index-file
* added missing *.prl files
* added mkdependHS as alternative
* new CVS version as of 1999/06/30
* Initial Release.