Simon Marchi via cfarm-users <cfarm-users@lists.tetaneutral.net> writes:
> On 2025-02-18 13:37, Bruno Haible via cfarm-users wrote: >> Martin Guy wrote: >>> $ ssh cfarm119 'echo $PATH' >>> /usr/bin:/etc:/usr/sbin:/usr/ucb:/usr/bin/X11:/sbin:/usr/java8_64/jre/bin:/usr/java8_64/bin >>> >>> Could that be added to ssh's default PATH? Then there'd be curl everywhere. >> >> Please not. It is important that the cfarm machines represent, as much >> as possible, the state of an arbirary machine with the same OS out there. >> >> It would invalidate our portability testing of package XY if the >> configure script of that package finds that all nice utitilies, >> from 'curl' to 'gperf', are available on AIX. Then we make a release >> of that package with that assumption. Then we get reports from people >> that on _their_ AIX machines that this assumption is not true. >> >> You are free to change you default PATH via .profile or .bashrc. >> But don't force it on all users of the cfarm. > > Agreed. > > I think it would be helpful to have MOTD messages indicating the > specifities of each machine, for instance where to find that magic > directory where the modern stuff is installed. In fact, I think this approach should actually be the documented intended way the cfarm machines should be setup. One of the first question that arise when building a package on exotic machine is what kind of additional software was installed on it compared to verbatim OS distribution. It is nice that a lot of GNU software is commonly available, but as Bruno said having it installed by default really hurts the purpose of build testing. Alas even /usr/local/bin is not safe since some software looks for additional components there by default. /Simon
signature.asc
Description: PGP signature
_______________________________________________ cfarm-users mailing list cfarm-users@lists.tetaneutral.net https://lists.tetaneutral.net/listinfo/cfarm-users
