[Bug 202633] Re: Please sync cernlib 2006.dfsg.2-11 (universe) from Debian unstable (main)
Cesare Tirabassi
cesare.tirabassi at gmail.com
Fri Mar 21 19:14:15 GMT 2008
cernlib (2006.dfsg.2-11ubuntu1) hardy; urgency=low
.
* Sync from Debian unstable (LP: #201962):
- Desktop file dropped in agreement with Debian (see #471039)
* debian/control and debian/control.d/0base.control:
- Change Build-Depends on gfortran to (>= 4:4.2.3) to avoid ftbfs
Once gfortran (>= 4:4.3) is available in Ubuntu this can be dropped.
* Modify Maintainer value to match Debian-Maintainer-Field Spec
.
cernlib (2006.dfsg.2-11) unstable; urgency=low
.
* Patch 102: Add kernlib/kerngen/tcgen/tlsc.F to the list of files not
to compile optimized, due to mis-compiled code at -O2 on ia64 (bug filed
against gfortran-4.3 as #466948). Should fix FTBFS on ia64.
* New patch 217: Force mathlib/gen test program "gent" to exit non-zero
if any individual test fails, so I don't have to inspect test suite
output manually for test failures in buildd logs.
.
cernlib (2006.dfsg.2-10) unstable; urgency=low
.
* Patch 304: Adjustments to compiler flags to solve various problems:
- Do not build with -funroll-loops, in order to work around compiler
bugs: ICE on arm at kernlib/kerngen/tcgen/ranecu.F (filed as #466791)
and bad code on ia64 at mathlib/gen/tests/c201s.F (filed as #466911).
- Build with -fno-range-check option to gfortran to ignore arithmetic
overflow compiler errors, as g77 does by default (closes: #466834),
and therefore...
* Patch 321: Revert work-around for that purpose from release -8.
.
cernlib (2006.dfsg.2-9) unstable; urgency=low
.
* Upload gfortran-transitioned CERNLIB to unstable.
.
cernlib (2006.dfsg.2-8) experimental; urgency=low
.
* Patch 321: In kernlib/kerngen/tcgens/lnxppcgs/{ubunch.F,ublow.F}, work
around gfortran overflow error in initializing 4-byte integer with
"data mask1/x'ff000000'/" statement. N.B. these files are only compiled
by big-endian architectures, so this change is irrelevant to i386, etc.
* debian/debhelper/cernlib-base.README.debian: Update to mention the new
montecarlo-data binary package and the soversion bump for libisajet.
* debian/control.d/libmathlib2-dev.control: Revert back to depending on
libblas-dev and liblapack-dev by default, as the libatlas-base-dev
package (and friends) does not ship liblapack.so and libblas.so symlinks
where the compiler and linker will see them automatically.
* debian/control.d/libmathlib2-dev.control: Also Conflict against
g77-based libdevel packages refblas3-dev and lapack3-dev.
* debian/NEWS: s/unstable/experimental/ in 2006.dfsg.2-7 NEWS entry to fix
newly introduced lintian warning.
.
cernlib (2006.dfsg.2-7) experimental; urgency=low
.
* Tweak dependencies and build-deps to use new libblas-dev and
liblapack-dev packages; also gfortran 4.3. No longer force installation
of BLAS and LAPACK reference implementations instead of ATLAS at
install time, only at build time.
.
* Patch updates to fix new issues that arose with gfortran 4.3:
- Patch 216: Fix compiler error "Result of LGAMMA overflows its kind" on
mathlib GEN test 310 by using the CERNLIB log-gamma function, not
the glibc one. Also force use of CERNLIB version of log-gamma function
internally elsewhere in mathlib.
- Patch 321: Fix ZEBRA test zebfz8 read error at EOF, caused by a
change in the definition of gfortran error codes between 4.2 and 4.3.
For some inexplicable reason, the error code returned by IOSTAT=xxx in
a ZEBRA call to READ at EOF is not the expected (-1), but is instead
the value defined as (LIB)ERROR_BAD_OPTION (3 in gfortran-4.2, 5002 in
gfortran-4.3) in libgfortran.h.
.
* Make packages installable in parallel with old g77-based runtime libs
(though NOT with old libdevel packages):
- Patch 605, debian/rules: Move kuesvr executable and man page elsewhere
so that libpacklib1-gfortran doesn't overwrite libpacklib1 files.
- Patch 800, debian/add-ons/cernlib.mk, debian/debhelper/lib*.install:
Add Make variable $(TOOL_NAME) to make shared lib filenames (as well
as sonames) include "_gfortran" string.
- debian/control.d/lib*.control: Remove relevant Conflicts/Replaces that
were introduced in previous revision.
- debian/NEWS: Edit most recent entry appropriately.
.
* Slight update to debian/README.64-bit to mention the -fno-automatic flag.
* Change install-cernlib-dirs.sh to link to Monte Carlo data files in
/usr/share/montecarlo-data (data files will be moved there in next
upload of mclibs to experimental).
* Debconf translation:
- Updated ja.po, thanks to Hideki Yamane <henrich at debian.or.jp>
(closes: #463710)
.
cernlib (2006.dfsg.2-6) experimental; urgency=low
.
* Experimental upload implementing gfortran transition:
- debian/add-ons/cernlib.mk: Make gfortran the default compiler.
- debian/add-ons/cernlib.mk: Add "_gfortran" string into sonames to keep
binary compatibility with Fedora gfortran-compiled CERNLIB packages.
- debian/add-ons/bin/cernlib.in: s/-lg2c/-lgfortran/g
- debian/NEWS: Warn that recompilation of user programs will be required.
- debian/README.amd64: Remove this file as it applies only to g77.
- debian/control.d/*, debian/debhelper/*, debian/copyright*, debian/rules:
Rename library packages for gfortran transition. While renaming
things, rename libpacklib1-lesstif{,-dev} to
libpacklib-lesstif1-{gfortran,dev} to get the soversion in the right
place in the package name. Add Conflicts and Replaces on old library
packages. Bump Build-Depends and Depends of devel packages to the first
versions of dependent packages built with gfortran.
.
* gfortran-related source code patch updates and revisions:
- Patch 102: Build gamma64.F with no compiler optimization; it fails
the mathlib test suite (GEN test #66, for E408) at -O3 in gfortran-4.2.
(Note, even when compiled optimized, the "failure" is caused by an
error in the 16th significant base-10 digit of a calculation involving
results returned from the DGAMMA function; I suspect mathlib authors
were overly optimistic about the precision of "double" variables.)
- Patch 215: Removed; hope the g77 compiler bug on ia64 that prompted this
patch isn't also present and needing to be worked around in gfortran.
- New patch 216: Force use of CERNLIB gamma function within mathlib, not
glibc's gamma function, which may be insufficiently precise; see, e.g.,
the thread started at http://gcc.gnu.org/ml/fortran/2007-10/msg00096.html
(This is not an issue now, but may be in the future when gfortran-4.3
becomes the default compiler.)
- Patch 304: Make sure that gfortran is passed the -fno-automatic flag.
- Patch 321: Rename MYGETARG to KERNLIBGETARG, and comment out GETARG
itself, to make name clashes much less likely. Use KERNLIBGETARG
instead of _gfortran_getarg_i4 in kfor.h in case gfortran name mangling
changes in the future.
.
* Support ppc64 architecture:
- Patch 211: No change needed; Patrice Dumas finds that like amd64, ppc64
puts static variables and local functions in memory under the 32-bit
barrier.
- Patch 304: Update Imake.cf to the version from xutils-dev 1:7.2.ds2-1
(but keeping the local change of darwin.cf -> MacOSX.cf). Revise
linux.cf to match, supplying -DCERNLIB_QMLXIA64 flag in the event
of Ppc64Architecture being defined.
.
* debian/cernlib-debian.mk: Remove section that would by default delete
the build tree of object files after stampdir/cernlib-{arch,indep}-stamp
were created. This could have caused a race condition when used with
parallel make. Conserving disk space on buildds is no longer as
important with CERN libs split into four separate source packages.
* debian/README.source: Remove mention of the now-useless "keepbuild"
$DEB_BUILD_OPTIONS flag.
.
* More lintian fixes:
- debian/control.d/0base.control: Bump Standards-Version to 3.7.3.
- debian/patches/*: Make sure all patches have a "## DP:" comment header.
- debian/rules: Add lintian override for 'cernlib' metapackage lintian
error in short description "description-starts-with-package-name".
.
* Miscellaneous minor tweaks:
- debian/cernlib-debian.mk: Define CERNLIB_VERBOSE to get more output
from some test suite programs.
- debian/control.d/0base.control: Add Homepage field.
- debian/copyright*: Mention the switch to GPL v3.
- debian/debhelper/kxterm.menu: Deleted; it is not useful to have kxterm
in the menu without an application to drive it.
- debian/debhelper/cernlib-base.README.debian: Explain how to turn on
Vim syntax highlighting for KUIPC and kumacs files, now that it is
no longer automatic.
- debian/*: Uniformly capitalize CERNLIB and CERN, following usage on
upstream's web page (but without editing old changelog or NEWS entries).
.
* debian/rules: Add rule to call debconf-updatepo explicitly from
debian/rules clean target.
* debian/control.d/*: debian/control reviewed by the debian-l10n-english
team as part of the Smith Review Project. (Partly closes: #454572)
A few additional tweaks made.
* debian/pawserv.templates: Debconf template reviewed by the debian-l10n-
english team as part of the Smith Review Project. (closes the rest of
#454572)
* New or updated debconf translations following the template update:
- Updated cs.po, thanks to Miroslav Kure <kurem at debian.cz>
(closes: #457682)
- Updated de.po, thanks to Erik Schanze <eriks at debian.org>
(closes: #457798)
- Updated fr.po, thanks to Christian Perrier <bubulle at debian.org>
(closes: #455775)
- Updated gl.po, thanks to Jacobo Tarrio <jtarrio at debian.org>
(closes: #455776)
- Updated pt.po, thanks to Miguel Figueiredo <elmig at debianpt.org>
(closes: #458597)
- Updated vi.po, thanks to Clytie Siddall <clytie at riverland.net.au>
(closes: #457302)
- New eu.po, thanks to Aitor Ibañez <aitiba at gmail.com>
(closes: #457829)
- New fi.po, thanks to Esko Arajärvi <edu at iki.fi>
(closes: #455962)
- New it.po, thanks to Luca Monducci <luca.mo at tiscali.it>
(closes: #456552)
- New nb.po, thanks to Bjørn Steensrud <bjornst at skogkatt.homelinux.org>
(closes: #455794)
- New ru.po, thanks to Yuri Kozlov <kozlov.y at gmail.com>
(closes: #434232)
** Changed in: cernlib (Ubuntu)
Importance: Undecided => Wishlist
Status: Confirmed => Fix Released
--
Please sync cernlib 2006.dfsg.2-11 (universe) from Debian unstable (main)
https://bugs.launchpad.net/bugs/202633
You received this bug notification because you are a member of Ubuntu
Package Archive Administrators, which is a direct subscriber.
More information about the ubuntu-archive
mailing list