IRC channel logs
2026-09-03.log
back to list of logs
<azeem>youpi: I'm wondering whether the recent libc fix for makeauth / setreuid might fix the "perl programs fail to send SIGHUP to processes of the same user but which were started by a different user" I reported in february <azeem>11:17 < azeem> right, the perl code runs setregid(105, -1); setreuid(101, -1); setresuid(-1, 101, -1) on Linux <azeem>hrm yeah, probably. It was a pretty weird corner-case in the way the Debian perl Postgres wrapper was switching uids, but I guess it boils down to that, yeah <azeem>youpi: but no, isn't the case of the above the X runs effectively as root, and the original user cannot send the signal? <azeem>my case is postgres being started as root but then set to user postgres, and then postgres cannot send the signal <youpi>but it's essentially the question of "non-user" <azeem>the user currently reported by ps is not able to send a signal ehre <azeem>if run that as sudo, then switch to user postgres (or uid 101 in this case) I cannot send HUP <youpi>switching to uid 101 and gid 105 <youpi>euid: 101, uid: 101, gid: 105 0 <youpi>$ ./testrun.sh /usr/bin/perl /tmp/test.pl <youpi>switching to uid 101 and gid 105 <youpi>Could not change user id at /tmp/test.pl line 14. <youpi>ah, no, I don't have the patch installed in that glibc build tree <youpi>-bash: kill: (22371) - Opération non permise <youpi>euid: 101, uid: 101, gid: 105 0 <youpi>I'm getting that in both cases <azeem>too bad, thanks for checking <etno>gnu_srs1: hi ! Any info to share about the above query ? ^^ <gnu_srs1>youpi: Maybe you should give back the build of python3.15. How to do that? BTW: We had a large power loss, but everything is OK by now. <youpi>gnu_srs1: yes, that's part of my buidld restart process :)