On Wed, 23 Apr 2025 08:47:08 +0200, Theo Buehler <t...@theobuehler.org> wrote: > > On Tue, Apr 22, 2025 at 10:29:21PM +0200, Kirill A. Korinsky wrote: > > On Tue, 22 Apr 2025 09:17:31 +0200, > > Theo Buehler <t...@theobuehler.org> wrote: > > > > > > Also, adding a pledge without giving a clear root cause is never the > > > right thing to do. At a minimum a ktrace should be provided. > > > > > > Why does this violate the inet pledge? Why does this work on 7.6? What > > > exactly changed in 7.7 so that it no longer works? > > > > > > These are questions that should be asked answered *before* commit. > > > > > > > Thanks for challenge it. > > > > Indeed, my patch was wrong and here no need to touch pledge(). > > > > The cause is claudio@ works on libutil's imsg. Because it includes renaming > > of imsg_init -> imsgbuf_init, it leads configure to a conclusion that tmate > > needs compat/imsg.c and compat/imsg-buffer.c where additional socket() is > > used. > > Yet, that makes sense. Thanks for digging into it. > > > So, here a diff which revert pledge() changes, and improved compatibility > > with new imsg API. > > > > Ok? > > ok, thanks! >
Am I right that we're agree that it's woth to be backported to 7.7? -- wbr, Kirill
