Re: bootstrap failure: libstdc++-v3/include/stdc++.h:83

[Jan Hubicka <jh at suse dot cz>]

> On Wed, 2003-07-09 at 04:10, Gerald Pfeifer wrote:
> > I don't know what might be causing this, but might this be due to one
> > of the large changes by you (Mark, Jan) in the last 18 hours?
> > 
> > Given that nobody else reported this, it might be i386 specific (as
> > opposed to i686 etc.).
> > 
> > /files/pfeifer/OBJ-0709-1215/gcc/xgcc -shared-libgcc -B/files/pfeifer/OBJ-0709-1215/gcc/ -nostdinc++  -L/files/pfeifer/OBJ-0709-1215/i386-unknown-freebsd4.8/libstdc++-v3/src -L/files/pfeifer/OBJ-0709-1215/i386-unknown-freebsd4.8/libstdc++-v3/src/.libs -B/sw/gcc-current/i386-unknown-freebsd4.8/bin/ -B/sw/gcc-current/i386-unknown-freebsd4.8/lib/ -isystem /sw/gcc-current/i386-unknown-freebsd4.8/include -isystem /sw/gcc-current/i386-unknown-freebsd4.8/sys-include -Winvalid-pch -Wno-deprecated -x c++-header -g -O2 -nostdinc++ -I/files/pfeifer/OBJ-0709-1215/i386-unknown-freebsd4.8/libstdc++-v3/include/i386-unknown-freebsd4.8 -I/files/pfeifer/OBJ-0709-1215/i386-unknown-freebsd4.8/libstdc++-v3/include -I/sw/test/gcc/cvs/libstdc++-v3/libsupc++ /sw/test/gcc/cvs/libstdc++-v3/include/stdc++.h -O2 -g -o ./i386-unknown-freebsd4.8/bits/stdc++.h.gch/O2g;
> > /sw/test/gcc/cvs/libstdc++-v3/include/stdc++.h:83: internal compiler error: Segmentation
> >    fault
> > Please submit a full bug report,
> > with preprocessed source if appropriate.
> > See <URL:http://gcc.gnu.org/bugs.html> for instructions.
> > /sw/test/gcc/cvs/libstdc++-v3/include/stdc++.h:83: internal compiler error: Segmentation
> >    fault
> > Please submit a full bug report,
> > with preprocessed source if appropriate.
> > See <URL:http://gcc.gnu.org/bugs.html> for instructions.
> > gmake[4]: *** [i386-unknown-freebsd4.8/bits/stdc++.h] Error 1
> > gmake[4]: Leaving directory `/files/pfeifer/OBJ-0709-1215/i386-unknown-freebsd4.8/libstdc++-v3/include'
> > gmake[3]: *** [all-recursive] Error 1
> > gmake[3]: Leaving directory `/files/pfeifer/OBJ-0709-1215/i386-unknown-freebsd4.8/libstdc++-v3'
> > gmake[2]: *** [all-recursive-am] Error 2
> > gmake[2]: Leaving directory `/files/pfeifer/OBJ-0709-1215/i386-unknown-freebsd4.8/libstdc++-v3'
> > gmake[1]: *** [all-target-libstdc++-v3] Error 2
> 
> I can't reproduce that bug, unfortunately -- at least not yet.
> 
> I've seen errors like that when I updated the compiler and didn't
> rebuild the PCH file -- was this build done in a completely clean tree? 
> (If this can be reproduced on the dbai.tuwien.ac.at machines, I can try
> to debug it there.)
I've reporduced this after configuring for i386-unkonwn-linux

The backtrace is somewhat weird:
#0  0x08355275 in ggc_get_size (p=0x0) at ../../gcc/ggc-page.c:542
#1  0x0820af45 in gt_pch_note_object (obj=0x2, note_ptr_cookie=0x2,
    note_ptr_fn=0x807f290 <gt_pch_p_14lang_tree_node>)
        at ../../gcc/ggc-common.c:274
#2  0x0807df7f in gt_pch_nx_lang_tree_node (x_p=0x2) at
gtype-cp.h:515
#3  0x0807e5bc in gt_pch_nx_lang_tree_node (x_p=0x1) at
gtype-cp.h:674
#4  0x0807e5bc in gt_pch_nx_lang_tree_node (x_p=0x1) at
gtype-cp.h:674
#5  0x0807e0dd in gt_pch_nx_lang_tree_node (x_p=0x41b30c80) at
gtype-cp.h:746
#6  0x0807e5bc in gt_pch_nx_lang_tree_node (x_p=0x1) at
gtype-cp.h:674
#7  0x0807e5bc in gt_pch_nx_lang_tree_node (x_p=0x0) at
gtype-cp.h:674
#8  0x0807e5bc in gt_pch_nx_lang_tree_node (x_p=0x0) at
gtype-cp.h:674
#9  0x0807e5bc in gt_pch_nx_lang_tree_node (x_p=0x1) at
gtype-cp.h:674
#10 0x0807e5bc in gt_pch_nx_lang_tree_node (x_p=0x0) at
gtype-cp.h:674
#11 0x0807e5bc in gt_pch_nx_lang_tree_node (x_p=0x1) at
gtype-cp.h:674
#12 0x0807e5bc in gt_pch_nx_lang_tree_node (x_p=0x0) at
gtype-cp.h:674
#13 0x0807e5bc in gt_pch_nx_lang_tree_node (x_p=0x1) at
gtype-cp.h:674
#14 0x0807e5bc in gt_pch_nx_lang_tree_node (x_p=0x0) at
gtype-cp.h:674
#15 0x0807e5bc in gt_pch_nx_lang_tree_node (x_p=0x0) at
gtype-cp.h:674
#16 0x0807e7b6 in gt_pch_nx_lang_tree_node (x_p=0x41ac0360) at
gtype-cp.h:589
#17 0x0807e882 in gt_pch_nx_lang_tree_node (x_p=0x41ab957c) at
gtype-cp.h:566
#18 0x0807e3fc in gt_pch_nx_lang_tree_node (x_p=0x41ae3c44) at
gtype-cp.h:667
#19 0x0807ebf0 in gt_pch_nx_lang_decl (x_p=0x41aa2cb8) at
gtype-cp.h:764
#20 0x0807e09e in gt_pch_nx_lang_tree_node (x_p=0x41ab94a4) at
gtype-cp.h:592

The first sensible looking x_p is identifier node. The other is function
and it gets into the zeros in decl_saved_tree traversal.  I guess these
are result of tail call optimization, so I will try to build unoptimized
cc1plus and see whether it still dies, but I will know the result not
before the evening.

Honza