On Thu, Sep 19, 2024 at 05:31:51PM +0100, Simon McVittie wrote: > Source: libunwind > Version: 1.7.2-1 > Severity: serious > Tags: ftbfs > Justification: fails to build from source (but built successfully in the past) > User: debian...@lists.debian.org > Usertags: i386 > > https://buildd.debian.org/status/fetch.php?pkg=libunwind&arch=i386&ver=1.7.2-1&stamp=1726720970&raw=0 > > x86/Gos-linux.c: In function '_Ux86_local_resume': > > x86/Gos-linux.c:302:22: error: passing argument 1 of '_Ux86_sigreturn' from > > incompatible pointer type [-Wincompatible-pointer-types] > > 302 | x86_sigreturn (sc); > > | ^~ > > | | > > | struct sigcontext * > > In file included from x86/Gos-linux.c:26: > > x86/unwind_i.h:64:42: note: expected 'unw_cursor_t *' {aka 'struct > > unw_cursor *'} but argument is of type 'struct sigcontext *' > > 64 | extern void x86_sigreturn (unw_cursor_t *cursor); > > | ~~~~~~~~~~~~~~^~~~~~ > > Possibly related to pre-existing bug #994510, which is also about uses > of struct sigcontext on i386.
Not much related. Both the arm64 and i386 FTBFS are "FTBFS with gcc 14", both are also present in the version currently in trixie, and both are fixed in 1.8.1. > I notice that many of the packages that link to libunwind on my system > seem to be using it as a nice-to-have feature to try to show backtraces > if they crash (on architectures supported by libunwind), rather than as > something that is functionally necessary. Should packages in this situation > be moving away from enabling libunwind on i386? Looking at number of test failures, armel/armhf/ppc64el seem to be currently the most broken release architectures. > Thanks, > smcv cu Adrian
