Segfault in erw64n_rebuild_gnu_hash

[BastianBlokland]

Thank you for the awesome tool, I'm facing a segfault when targeting glibc 2.31 on a executable built with clang-18 (and linked with lld).

Starting program: /home/bastian/dev/external/polyfill-glibc/polyfill-glibc volo-polyfill-repro --target-glibc=2.31

Program received signal SIGSEGV, Segmentation fault.
0x00005555555bad66 in erw64n_rebuild_gnu_hash (erw=0x7fffffffd760) at src/erw_nne.h:1810
1810	      buckets[i] = 0;
(gdb) p i
$1 = 932924
(gdb) bt
#0  0x00005555555bad66 in erw64n_rebuild_gnu_hash (erw=0x7fffffffd760) at src/erw_nne.h:1810
#1  0x00005555555bb98a in erw64n_dsyms_flush (erw=0x7fffffffd760) at src/erw_nne.h:1884
#2  0x00005555555c22b3 in erw64n_flush (erw=0x7fffffffd760) at src/erw_nne.h:2574
#3  0x00005555555f9289 in erw_flush (erw=0x7fffffffd760) at src/erw.c:1328
#4  0x000055555556e530 in main (argc=3, argv=0x7fffffffdad8) at src/main.c:972

It seems that old_u32_count in erw_nne.h:1750 is invalid: 4294967198.

Repro:

wget https://bastian.tech/temp/volo-polyfill-repro.gz
gunzip volo-polyfill-repro.gz
./polyfill-glibc volo-polyfill-repro --target-glibc=2.31

[BastianBlokland]

Cause seems to be erw->dsyms.original_gnu_hash_count being 0 while erw->dsyms.original_had_gnu_hash is true.

[BastianBlokland]

Checking for erw->dsyms.original_had_gnu_hash being non-zero on line 1746 seems to work correctly (at least the processed executable seems to run correctly), but I don't know enough about the implementation to say if that is a good patch.