build fails on i386

Bug #6511 reported by Allison Karlitskaya
10
Affects Status Importance Assigned to Milestone
clisp
Invalid
Undecided
Unassigned
clisp (Ubuntu)
Fix Released
Medium
StefanPotyra

Bug Description

The current version of clisp is missing from the archives for 386. This is caused by a build failure.

To correct the build failure -falign-functions=4 needs to be added to the Makefile.

Changed in clisp:
assignee: nobody → motu
Lukas Fittl (lfittl)
Changed in clisp:
status: Unconfirmed → Confirmed
Revision history for this message
Peter Van Eynde (ubuntu-pvaneynd) wrote :

As the debian maintainer of this package I would like to know what version you are refering to.

Please note that I have not been informed of this problem and that clisp 2.36-X builds correctly on i386 on debian.

Revision history for this message
Stephan Rügamer (sruegamer) wrote :

Hmmm...

if Ryan is meaning this, what happened to me, when I tried to build clisp, that you'll find it in the attached buildlog.

Revision history for this message
Stephan Rügamer (sruegamer) wrote : buildlog of clisp_2.36-3 on dapper (using i386 chroot on an amd64 box)

attached buildlog

Revision history for this message
Peter Van Eynde (ubuntu-pvaneynd) wrote :

What I see in the buildlog is a failure of the x86_64 (so amd64) code. Please not that in debian we found an ICE in gcc on amd64, see http://bugs.debian.org/cgi-bin/bugreport.cgi?bug=349268 .

So:
- does the -falign-functions=4 fix the ICE on amd64?
- or is it a bug because we try to build the amd64 stuff in the
  i386 chroot on an amd64 machine?

Revision history for this message
Stephan Rügamer (sruegamer) wrote :

Well,

I will check the clisp build on plain amd64.
If it failes again, I will try to add -falign-functions=4 to the CFLAGS (is it enough to add it to the CFLAGS during configure, or do I need to tweak the autotools magic?)

If I faile or succeed, I'll report it here.

Revision history for this message
Stephan Rügamer (sruegamer) wrote :

Ok,

first of all, if I build clisp on amd64, it compiles the amd64 stuff, but fails later, and this is, IMHO, your gcc4 bug.
So the i386 build failures are coming from trying to build amd64 stuff on i386.

Regards,

\sh

CLISP="`pwd`/lisp.run -M `pwd`/lispinit.mem -B `pwd` -N `pwd`/locale -Efile UTF-8 -Eterminal UTF-8 -Emisc 1:1 -norc" ; cd syscalls ; dots=`echo syscalls/ | sed -e 's,[^/][^/]*//*,../,g' -e 's,/$,,g'` ; make clisp-module CC="gcc" CPPFLAGS=" " CFLAGS="-W -Wswitch -Wcomment -Wpointer-arith -Wimplicit -Wreturn-type -Wmissing-declarations -Wno-sign-compare -O -DUNICODE -DDYNAMIC_FFI -DDYNAMIC_MODULES -I. -fPIC" INCLUDES="$dots" CLFLAGS="-x none -Wl,-export-dynamic" LIBS="libcharset.a libavcall.a libcallback.a /usr/lib/libreadline.so -lncurses -ldl -L/usr/lib -lsigsegv" RANLIB="ranlib" CLISP="$CLISP -q"
make[2]: Entering directory `/tmp/buildd/clisp-2.36/debian/build/syscalls'
/tmp/buildd/clisp-2.36/debian/build/lisp.run -M /tmp/buildd/clisp-2.36/debian/build/lispinit.mem -B /tmp/buildd/clisp-2.36/debian/build -N /tmp/buildd/clisp-2.36/debian/build/locale -Efile UTF-8 -Eterminal UTF-8 -Emisc 1:1 -norc -q ../modprep.fas calls.c
;; MODPREP: "calls.c" --> #P"calls.m.c"
;; MODPREP: reading "calls.c": 143,149 bytes, 3,796 lines
;; MODPREP: 476 objects, 72 DEFUNs
;; packages: ("OS" "POSIX")
MODPREP: wrote calls.m.c (396,547 bytes)
gcc -W -Wswitch -Wcomment -Wpointer-arith -Wimplicit -Wreturn-type -Wmissing-declarations -Wno-sign-compare -O -DUNICODE -DDYNAMIC_FFI -DDYNAMIC_MODULES -I. -fPIC -I.. -c calls.m.c -o calls.o
calls.c: In function 'get_block':
calls.c:843: error: unrecognizable insn:
(insn 173 172 174 0 (set (reg/f:DI 91)
        (plus:DI (reg:DI 133)
            (const_int 1125899906842624 [0x4000000000000]))) -1 (insn_list:REG_DEP_TRUE 172 (nil))
    (expr_list:REG_DEAD (reg:DI 133)
        (expr_list:REG_EQUAL (const:DI (plus:DI (symbol_ref:DI ("symbol_tab_data") [flags 0x40] <var_decl 0x2aaaab25f820 symbol_tab_data>)
                    (const_int 1125899906842624 [0x4000000000000])))
            (nil))))
calls.c:843: internal compiler error: in extract_insn, at recog.c:2020
Please submit a full bug report,
with preprocessed source if appropriate.
See <URL:http://gcc.gnu.org/bugs.html> for instructions.
For Debian GNU/Linux specific bug reporting instructions,
see <URL:file:///usr/share/doc/gcc-4.0/README.Bugs>.
make[2]: *** [calls.o] Error 1
make[2]: Leaving directory `/tmp/buildd/clisp-2.36/debian/build/syscalls'
make[1]: *** [syscalls] Error 2
make[1]: Leaving directory `/tmp/buildd/clisp-2.36/debian/build'
make: *** [build-stamp] Error 2
pbuilder: Failed autobuilding of package

Revision history for this message
Dave Love (fx-gnu) wrote :

The reason the build fails is that rules doesn't specify
gcc-3.3 for i386 and gcc-4 is being used. The latest Debian
(2.38-1) source builds OK with the following change.

Revision history for this message
Dave Love (fx-gnu) wrote : fix for i386 build

Fix referred to in the previous comment.

[Sorry I don't understand how to drive this reporting
system. Oh for email.]

Revision history for this message
Dave Love (fx-gnu) wrote : correct fix

Sorry the previous diff was bogus. Double oh for email.
This is what I meant to send.

Revision history for this message
StefanPotyra (sistpoty) wrote :

Hi,

(You'd need to udpate the build-depends needing to include gcc-3.3 for i386 then as well), however we can't compile packages with gcc-3.3 any longer, as it might get removed soon. I'll recheck this with doko, maybe he's got a clue about the gcc-issue.

Cheers,
     Stefan.

StefanPotyra (sistpoty)
Changed in clisp:
assignee: motu → sistpoty
Revision history for this message
Dave Love (fx-gnu) wrote : alternative patch

Ah,obviously I was confused by the build-deps
somehow, not noticing the architecture dependence.
Anyway if you want to eliminate gcc-3.3, this will fix it
for i386. You'll also have to fix that clisp-dev
depends (not build-depends) on gcc-3.3.

Revision history for this message
StefanPotyra (sistpoty) wrote :

Hi,

Thx. for your help and comments, I've just uploaded a version. I'm pretty excited to see the amd64 buildlog ;).

Cheers,
     Stefan.

Changed in clisp:
status: Confirmed → Fix Committed
Revision history for this message
StefanPotyra (sistpoty) wrote :

Thanks for all your comments. It took me only 4 uploads *cough* to find out, that sparc/amd64 will *only* succeed with gcc-3.3 and that Dave's patch works (actually used for all arches), applied to debian's 1:2.38-1 package of clisp.

Finally closing this bug ;).

@Peter: I did a very small patch with dh_gencontrol magic to eliminate gcc-3.3 dependency from clisp-dev for arches which don't get build with gcc-3.3. If you want, I'll sort it out and put it to debian bts.

Cheers,
    Stefan.

Changed in clisp:
status: Fix Committed → Fix Released
Revision history for this message
Peter Van Eynde (ubuntu-pvaneynd) wrote :

Thanks, patches integrated into debian package.

Changed in clisp:
status: Unconfirmed → Rejected
To post a comment you must log in.
This report contains Public information  
Everyone can see this information.

Other bug subscribers

Remote bug watches

Bug watches keep track of this bug in other bug trackers.