Differences between revisions 1 and 73 (spanning 72 versions)
Revision 1 as of 2005-12-20 14:09:16
Size: 3138
Editor: ?smarenka
Comment:
Revision 73 as of 2006-06-27 20:35:59
Size: 4146
Editor: ?smarenka
Comment:
Deletions are marked like this. Additions are marked like this.
Line 1: Line 1:
This page tracks current issues with GCC for DebianM68kPorting. User support is at debian-68k@lists.debian.org. Buildd admins can be found at m68k-build@nocrew.org. This page tracks current issues with ["GCC"] for DebianM68kPorting. User support is at debian-68k@lists.debian.org. Buildd admins can be found at m68k-build@nocrew.org.

Since gcc-4.1 is now the default, I'm going to start dropping bugs that only apply to older packages.
Line 5: Line 8:
 * If you find a bug not listed [http://bugs.debian.org/cgi-bin/pkgreport.cgi?tag=m68k;users=m68k-build@nocrew.org;pri0=tag:help,moreinfo,needs-help,needs-port,pending,upstream,patch,gcc-40,toolchain here], please usertag it or drop us a message.
 * S
ee DebianM68kPorting for more info about usertags.
 * If you find a bug not listed [http://bugs.debian.org/cgi-bin/pkgreport.cgi?tag=m68k;users=m68k-build@nocrew.org;pri0=tag:help,moreinfo,needs-help,needs-port,pending,upstream,patch,gcc-40,toolchain here], please usertag it or drop us a message, see DebianM68kPorting.
Line 9: Line 11:
= gcc-4.0 bugs = = gcc-4.x bugs =
Line 11: Line 13:
 * debian [http://bugs.debian.org/cgi-bin/pkgreport.cgi?src=gcc-4.0 bts]  * debian [http://bugs.debian.org/cgi-bin/pkgreport.cgi?src=gcc-4.0 gcc-4.0 bts]
 * debian [http://bugs.debian.org/cgi-bin/pkgreport.cgi?src=gcc-4.1 gcc-4.1 bts]
Line 15: Line 18:
== fixed (tracking packages) ==

 * ICE in cp_expr_size ('''fixed''')
  * debian #323686 and 323133, gnu pr 21123
  * found in gcc-4.0 4.0.2-2, fixed in 4.0.2-5.
  * rebuild: kfolding_1.0.0-rc2-4 klibido_0.2.4.1-1 ksubtile_1.1-1 ktorrent_1.1-1 ktrack_0.3.0-alpha1-6.1
  * done: aleph_0.9.0-2.1, ardour_0.99-3, bibletime_1.5.1-1, baghira_0.7a-1, inkscape_0.42.2+0.43pre1-1
Line 25: Line 21:
 * ICE in instantiate_virtual_regs_lossage
  * debian #333536, gnu pr 25043 (last updated 20051126)
  * from 25043: fixed in 4.1, 4.2, fails 4.0.3
  * found in gcc-4.0 4.0.2-5
  * pkgs: cl-sql_3.3.3-1, ntop_3:3.2-1, plptools_0.14-1, rsplib_1.0.0-1, squashfs_1:2.2r2-1, xfsprogs_2.7.7-1
 * csound_1:4.23f13-1: executable segfaults
  * debian #[http://bugs.debian.org/342121 342121], gnu pr [http://gcc.gnu.org/bugzilla/show_bug.cgi?id=25496 25496] (filed 20051219)
  * found in gcc-4.0 4.0.2-7, gcc-4.1 4.1-0exp4
  * note: accessing a char array as a long is an aliasing violation, per 25496 -- share with csound maintainer.
  * shared per debian #[http://bugs.debian.org/350757 350757]
  * won't be fixed by upstream, is now closed (05/2006)
Line 31: Line 28:
 * ICE: output_operand: invalid expression as operand:
  * debian #270340, gnu pr 18592 (last updated 20050723)
  * found in gcc-4.0 4.0.2-2
  * pkgs: gnutls11_1.0.16-14, openmsx_0.5.2-4, zynaddsubfx_2.2.1-2.1 log4cxx_0.9.7-4 (w/gcc-4.0)
  * Ask if this needs to be filed separately against 4.0.3? Check test case against 4.0.3.
 * make: fatal error: internal consistency failure
  * debian #[http://bugs.debian.org/344041 344041], gnu pr [http://gcc.gnu.org/bugzilla/show_bug.cgi?id=25514 25514]
  * found in gcc-4.0 4.0.2-7, gcc-4.1 4.1-0exp4
  * noted upstream that it's also found in 4.1
  * patch available (going into 4.1.2?)
Line 37: Line 34:
 * nmh: ICE reload_cse_simplify_operands
  * debian #340293, gnu pr 24912 (last updated 20051121)
  * from 24912: fixed in trunk? (gnu pr is closed)
  * found in gcc-4.0 4.0.2-5
  * Ask if this needs to be filed separately against 4.0.3? Check test case against 4.0.3.
 * cppunit: optimizer bug: registers are clobbered but not restored
  * debian #[http://bugs.debian.org/345574 345574], gnu pr [http://gcc.gnu.org/bugzilla/show_bug.cgi?id=25642 25642]
  * found in gcc-4.0 4.0.2-5, fixed in 4.0.2-7 or worked around in cppunit?
  * retry aptitude once cppunit is fixed
Line 43: Line 39:
 * csound_1:4.23f13-1: executable segfaults
  * debian #342121, gnu pr 25496 (filed 20051219)
  * found in gcc-4.0 4.0.2-5
  * note: accessing a char array as a long is an aliasing violation, per 25496 -- share with csound maintainer.
 * Assembler messages: Error: value out of range
  * debian #[http://bugs.debian.org/359281 359281], gnu pr [http://gcc.gnu.org/bugzilla/show_bug.cgi?id=26415 26415]
  * found in gcc-4.0 4.0.2-3, gcc-4.1.1-2
  * pkgs: torcs_1.2.4-1, grace_1:5.1.18-4 ([http://bugs.debian.org/338433 338433]), rscheme_0.7.3.4.b1-2.1 krecipes, synfig_0.61.05-5 parrot
  * works fine with latest binutils and gcc-3.3, but not gcc-4.0.
  * torcs & grace work with -O1 (both gcc-4.1 & gcc-4.0)
  * rscheme works with gcc-4.1 -O2 & gcc-4.0 -O1

 * error: insn does not satisfy its constraints
  * debian #[http://bugs.debian.org/375522 375522], gnu pr [http://gcc.gnu.org/PR28181 28181]
  * pkgs: nmh, openmsx_0.6.0-1?
  * found in gcc-4.1_4.1.1-2
Line 51: Line 55:
 * make: fatal error: internal consistency failure
  * debian #344041
  * found in gcc-4.0 4.0.2-5
 * gij-4.0: dies with sigabrt
  * debian #[http://bugs.debian.org/340874 340874]
  * found in gij-4.0 4.0.2-4
  * see also kaffe, ecj-bootstrap, gjdoc
  * test vs. 4.1
Line 55: Line 61:
 * fftw3: ICE in reload
  * debian #323426,
  * found in gcc-3.4 3.4.4-9, found in gcc-4.0 4.0.2-4
  * needs a test case
 * cmake: Illegal instruction
  * debian #[http://bugs.debian.org/360586 360586]
  * found in gcc-4.0 gcc-4.1
  * pkgs affected: vtk, octaviz
  * rebuilding cmake with -O0 allows vtk to build
  * test against gcc-4.1_4.1.1-2
  * related to self-test failures?

 * cln: assembler: FATAL: can't close cl_MI.o: no error
  * debian #[http://bugs.debian.org/375519 375519]
  * found in gcc-4.1_4.1.1-2
Line 62: Line 76:
 * asterisk_1:1.0.9.dfsg-5: segfault
  * found in gcc-4.0 4.0.2-3
 * segfaulting exe's
   * found in gcc-4.0 4.0.2-9
   * pkgs: drscheme ecl emacs-snapshot libgc libgstreamer-perl newmat openoffice.org pnet synfigstudio
   * Related to reopened [http://bugs.debian.org/331150 331150]?
   * gcc-4.1 is the default, retry these or whatever package created the segfaulting app
   * illegal instruction vs. segfault
Line 65: Line 83:
 * Assembler messages: Error: value out of range
  * found in gcc-4.0 4.0.2-3
  * pkgs: torcs_1.2.4-1, grace_1:5.1.18-4 (338433)
  * works fine with latest binutils and gcc-3.3, but not gcc-4.0.
  * needs a test case and a gcc-4.0 bug filed.
Line 71: Line 84:
== gcc-3.4 bugs ==
 
  * ICE verify_initial_elim_offsets
   * found in gcc-3.4 3.4.4-9
   * gcc-4.0 yields 323133 (fixed in 4.0.3)
   * pkgs: kdemultimedia, koffice
   * should build with gcc-4.0_4.0.2-5
   * kdemultimedia is confirmed to build with gcc-4.0_4.0.2-5.
== binNMUs ==

 * binNMU [http://bugs.debian.org/cgi-bin/pkgreport.cgi?tag=binNMU;users=m68k-build@nocrew.org bugs]
 * make (variable.c -O1) [http://bugs.debian.org/344046 344046]
  * needed for implicit make dependencies
 * aptitude (skip self-tests) [http://bugs.debian.org/353719 353719]
  * needed to make base installable for d-i
 * cmake (-O0) [http://bugs.debian.org/360586 360586]
  * needed for vtk and octaviz

This page tracks current issues with ["GCC"] for ?DebianM68kPorting. User support is at debian-68k@lists.debian.org. Buildd admins can be found at m68k-build@nocrew.org.

Since gcc-4.1 is now the default, I'm going to start dropping bugs that only apply to older packages.

m68k usertags

gcc-4.x bugs

upstream

not upstream yet

  • gij-4.0: dies with sigabrt
  • cmake: Illegal instruction
    • debian #[http://bugs.debian.org/360586 360586]

    • found in gcc-4.0 gcc-4.1
    • pkgs affected: vtk, octaviz
    • rebuilding cmake with -O0 allows vtk to build
    • test against gcc-4.1_4.1.1-2
    • related to self-test failures?
  • cln: assembler: FATAL: can't close cl_MI.o: no error

no gcc bug filed yet

  • segfaulting exe's
    • found in gcc-4.0 4.0.2-9
    • pkgs: drscheme ecl emacs-snapshot libgc libgstreamer-perl newmat openoffice.org pnet synfigstudio
    • Related to reopened [http://bugs.debian.org/331150 331150]?

    • gcc-4.1 is the default, retry these or whatever package created the segfaulting app
    • illegal instruction vs. segfault

binNMUs