Re: lib/59564: ELF source code compatibility regression

["Joseph Koshy" <netbsd%jkoshy.net@localhost>]

On 2025-12-15 14:50 David H. Gutteridge via gnats <gnats-admin%netbsd.org@localhost> wrote:

> /tmp/pkgsrc_work/www/firefox140/work/firefox-140.6.0/third_party/abseil-cpp/absl/debugging/internal/elf_mem_image.cc:385:45: 
>  error: request for member 'vs_vers' in '* version_symbol', which is of 
>  non-class type 'const Elf64_Versym' {aka 'const short unsigned int'}
>     385 |   const int version_index = version_symbol->vs_vers & 
>  VERSYM_VERSION;
>         |                                             ^~~~~~~
> 
>  I haven't bisected/analyzed/anything else, just noting this here.

Apologies for the breakage.

The definition of Elf{32,64}_Versym was changed in -CURRENT ([1])
to align its definition with other ELF toolkits (and with the
file format).

There are two compile-time guards that could be used to compile
around this change:

 a) a bump of __NetBSD_version__ ([2]),
 b) increment to an inclusion guard ([3]).

 [1]: https://freshbsd.org/netbsd/src/commit/tQg9ppOQUqMIebiG
 [2]: https://freshbsd.org/netbsd/src/commit/oRTT2fQq7NfhgjiG
 [3]: https://freshbsd.org/netbsd/src/commit/q8SBQRaMqO28zPiG

--
Joseph Koshy | jkoshy%NetBSD.org@localhost