Hi! On Sun, 2024-03-03 at 23:00:00 +0100, Guillem Jover wrote: > On Thu, 2024-02-29 at 02:35:16 +0100, Guillem Jover wrote: > > Control: tags -1 - pending > > > On Wed, 2024-01-31 at 19:36:09 +0000, Steve Langasek wrote: > > > Source: libaio > > > Version: 0.3.113-5 > > > Severity: serious > > > Tags: patch pending > > > Justification: library ABI skew on upgrade > > > User: debian-...@lists.debian.org > > > Usertags: time-t > > > > > Please find the patch for this NMU attached. > > > > > > If you have any concerns about this patch, please reach out ASAP. > > > Although > > > this package will be uploaded to experimental immediately, there will be a > > > period of several days before we begin uploads to unstable; so if > > > information > > > becomes available that your package should not be included in the > > > transition, > > > there is time for us to amend the planned uploads. > > > > Unfortunately I just realized this patch is not enough. :/ This library > > emits direct syscalls, so these are going to be broken with the time_t > > size change, the syscalls need to be updated. I'm checking how to best > > fix this, perhaps even via dual-ABI, to avoid the transition > > altogether, but let's see. > > > > I guess this might have been missed for other packages that that emit > > direct syscalls and are not using the time64 variants for those > > already. > > Just as a status update, I've got most of this working, but upstream > does not tend to be very responsive, so I think I'll do a proper > SONAME bump with my proposed changes for the dual-ABI, to avoid any > potential clashes with anything that gets upstream, and to make a > revert easier, by reusing the t64 library names. And then once/if this > gets merged upstream I can revert that and simply do the proper > dual-ABI on the old SONAME and package names, as if nothing had > happened (except for the required rebuilds). > > Hopefully I can have something for upload today or tomorrow, hoping > that this delay up to now, does not block too many things. :/
I've got all the upstream changes now ready, except that there's still one test case failing, something wrong with the sigset_t type. I've run out of time trying to track this down, but I've pushed what I have on the pu/time64 branch, and will continue later today. Thanks, Guillem
