[Bug 276614] graphics/mesa-dri:fix os_same_file_description warning
- In reply to: bugzilla-noreply_a_freebsd.org: "[Bug 276614] graphics/mesa-dri:fix os_same_file_description warning"
- Go to: [ bottom of page ] [ top of archives ] [ this month ]
Date: Wed, 27 Mar 2024 16:09:27 UTC
https://bugs.freebsd.org/bugzilla/show_bug.cgi?id=276614 Jan Beich <jbeich@FreeBSD.org> changed: What |Removed |Added ---------------------------------------------------------------------------- CC|jbeich@FreeBSD.org | --- Comment #44 from Jan Beich <jbeich@FreeBSD.org> --- (In reply to Yusuf Khan from comment #42) > although if the memory leak is as severe as you say on certain applications I've challenged your severity claim but didn't state my own, only pointed at related upstream findings. The warning disappeared from my dogfood plate 3 years ago in ports c14ad44ccafc, so I didn't have time to gather much data. The ongoing cost of uncertainty in my own QA and when reporting various bugs upstream was too much to not just fix it. (In reply to Ivan Rozhuk from comment #43) > 1. How you has detect mem leak? Indirectly by logging every (not just the first) failed os_same_file_description() call (with PID and errno) into /var/log/messages. This doesn't track the leaked memory size. > 2. How it is possible to leak in syslog()? Look inside the patch. It merely hooks ad hoc warnings into MESA_LOG facility and changes MESA_LOG from "file" to "syslog" by default. Logging into to stderr isn't reliable if Xorg is started by a greeter (graphical login manager aka display manager) and changing MESA_LOG_FILE isn't reliable with setuid applications such as Xorg (until bug 273161). Besides, Mesa warnings are kinda serious unlike the rest of stderr output after startx, which often includes regular and noisy applications. -- You are receiving this mail because: You are the assignee for the bug.