Segfault on dumping from CVS

[Mike Hearn]

Hi,

I can't build emacs from CVS on fedora core:

Dumping under names emacs and emacs-21.3.50
make[1]: *** [bootstrap-emacs] Segmentation fault

Program received signal SIGSEGV, Segmentation fault.
0x00ad9c4c in memcpy () from /lib/tls/libc.so.6
(gdb) bt
#0  0x00ad9c4c in memcpy () from /lib/tls/libc.so.6
#1  0x0818f8b6 in unexec (new_name=0x9ccff80 "/source/emacs/src/emacs", 
old_name=0x9ccff9c "/source/emacs/src/temacs",
    data_start=138060232, bss_start=0, entry_address=0) at unexelf.c:925
#2  0x0810da72 in Fdump_emacs (filename=164609820, symfile=164609804) at 
emacs.c:2155
#3  0x0816794e in Feval (form=136054888) at eval.c:2078
#4  0x081654ba in Fprogn (args=-1095531784) at eval.c:408
#5  0x08165409 in Fif (args=-1450257568) at eval.c:357
#6  0x08167a9f in Feval (form=-2009431140) at eval.c:2025
#7  0x081653e8 in Fif (args=-1450257272) at eval.c:356
#8  0x08167a9f in Feval (form=-2009431140) at eval.c:2025
#9  0x0817b057 in readevalloop (readcharfun=695816204, stream=0x97c1838, 
sourcename=1769706572, evalfun=0x8167614 <Feval>,
    printflag=0, unibyte=138062976, readfun=695703604) at lread.c:1376
#10 0x0817a138 in Fload (file=1769706572, noerror=695703604, 
nomessage=695703604, nosuffix=695703604, must_suffix=695703604)
    at lread.c:914
#11 0x0816799c in Feval (form=138056744) at eval.c:2092
#12 0x0810e959 in top_level_2 () at keyboard.c:1303
#13 0x081667fe in internal_condition_case (bfun=0x810e948 <top_level_2>, 
handlers=695797284, hfun=0x810e650 <cmd_error>)
    at eval.c:1333
#14 0x0810e985 in top_level_1 () at keyboard.c:1311
#15 0x0816639d in internal_catch (tag=138062976, func=0x810e95c <top_level_1>, 
arg=695703604) at eval.c:1094
#16 0x0810e8d7 in command_loop () at keyboard.c:1272
#17 0x0810e42c in recursive_edit_1 () at keyboard.c:989
#18 0x0810e53c in Frecursive_edit () at keyboard.c:1045
#19 0x0810ce99 in main (argc=5, argv=0xbfee42b4) at emacs.c:1666


(gdb) list
920                  new data2 section always be placed in the same spot as the 
old
921                  bss section by any other application. */
922               NEW_SECTION_H (nn).sh_addralign = OLD_SECTION_H 
(n).sh_addralign;
923
924               /* Now copy over what we have in the memory now. */
925               memcpy (NEW_SECTION_H (nn).sh_offset + new_base,
926                       (caddr_t) OLD_SECTION_H (n).sh_addr,
927                       new_data2_size);
928               nn++;
929             }


Has anybody seen this before?

thanks -mike