GCC g++ ICE on i586
Affects | Status | Importance | Assigned to | Milestone | |
---|---|---|---|---|---|
| Linaro GCC |
Fix Released
|
Low
|
Andrew Stubbs | |
| Linaro GCC Tracking |
In Progress
|
Undecided
|
Unassigned |
Bug Description
Attached testcase causes GCC to ICE when cross compiled for i586 to reproduce compile the attached file with
g++ -Os -march=pentium libc-tls.c
ios_init.c: In destructor 'std::ios_
ios_init.c:15899:3: internal compiler error: in
dwarf2out_
Please submit a full bug report,
with preprocessed source if appropriate.
See <http://
Related branches
Khem Raj (khem-raj) wrote : | #1 |
Khem Raj (khem-raj) wrote : | #2 |
Michael Hope (michaelh1) wrote : | #3 |
Confirmed in gcc-linaro-
Changed in gcc-linaro: | |
status: | New → Triaged |
importance: | Undecided → Low |
Michael Hope (michaelh1) wrote : | #4 |
Hits an assertion at the end of dwarf2out_
/* And emulate the state save. */
gcc_assert (!cfa_remember.
cfa_remember = cfa;
cfa_remember.
}
Set to low priority as it's size only in a non-code generation area.
Michael K. Edwards (m-k-edwards) wrote : | #5 |
This bug is still present in the 2011.03-0 milestone build. It breaks the build of eglibc's libstdc++ (and therefore the construction of a cross toolchain) at -mcpu=pentium-m -Os, which is the tuning we've been using for our embedded device.
Adding -fomit-
build@ctbu-bld5:~$ /home/build/
build@ctbu-bld5:~$ /home/build/
/home/build/
/home/build/
Please submit a full bug report,
with preprocessed source if appropriate.
See <http://
Michael K. Edwards (m-k-edwards) wrote : | #6 |
I of course did not mean eglibc's libsdc++; I mean Linaro gcc's. (I was *very* tired when I wrote that.) In case it matters, the failure occurred during the "final" compiler phase (after libc installation) of a toolchain build done with crosstool-ng, using an eglibc patched similarly to current Natty.
Michael Hope (michaelh1) wrote : | #7 |
Note that adding '-fno-shrink-wrap' works around the problem.
Michael Hope (michaelh1) wrote : | #8 |
LP: #730860 is related.
Michael K. Edwards (m-k-edwards) wrote : | #9 |
The patch attached to https:/
Andrew Stubbs (ams-codesourcery) wrote : | #10 |
I've just committed a patch from Bernd Schmidt. This issue should now be fixed.
Related: lp:gcc-linaro/4.5,revno=99486
Changed in gcc-linaro: | |
assignee: | nobody → Andrew Stubbs (ams-codesourcery) |
milestone: | none → 4.5-2011.04-0 |
status: | Triaged → Fix Committed |
Changed in gcc-linaro-tracking: | |
milestone: | none → 4.7.0 |
tags: | added: shrinkwrap |
Andrew Stubbs (ams-codesourcery) wrote : | #11 |
Bernd has now posted this upstream:
http://<email address hidden>
Changed in gcc-linaro-tracking: | |
status: | New → In Progress |
Michael Hope (michaelh1) wrote : | #12 |
This is fixed in more recent builds. Tested using:
michaelh@
on x86_64.
Changed in gcc-linaro: | |
status: | Fix Committed → Fix Released |
tags: | added: 46merge |
hmm I forgot to mention that if I revert r99474 then the ICE goes away. I can not reproduce this issue with upstream gcc trunk