On Thu, Aug 29, 2024 at 08:46:41AM GMT, Gábor Németh wrote: > > I would appreciate help typing in the additional code and submitting > > that to upstream, so we don't have to diverge from upstream forever. > > I did the typing-in part and have a small patch [1].
Wow, great work! > Before seding the PR I'd appreciate a quick sync with you. > > My approach was to simply add the second argument, making an sg symlink > trivially work. Possible objection may be that original newgrp did not > accept the command argument, thus `newgrp group some-cmd` would execute > some-cmd instead of ignoring it. Also upstream advises against adding > new short options [2] but I can argue that in case of `newgrp` we are > returning to some established behavior (sg), or that the cited SUSv2 is > not followed to the letter anyway as eg. it prescribes accepting '-' as > the group name, which upstream does not do. A solution complicating the > patch would be to test for argv[0] if we're called as `sg` but I found > that ugly and can't decide. Okay, I hear your conccerns. For sg, I think one can argue the value of providing it is in providing it with the existing behaviour from shadow, providing compatibility. A completely interface would certainly look differently. > I'd appreciate your thoughts. I've dug around a bit in the upstream sources, and it appears they like separate source files with their own main() function for each program name. But I'd advise sending the patch upstream and/or to the mailing list and seeing what they think! Best, Chris
