Re: RUN_PRECMD hook

[David Brownlee <abs%absd.org@localhost>]

On Sun, 12 Apr 2026 at 15:01, Jonathan Perkin <jperkin%pkgsrc.org@localhost> wrote:
>
> One final part of my bulk build migration is support for cputime limits
> to stop spinning processes from hanging builds indefinitely.  In pbulk I
> simply had this line in pbulk.conf:
>
>    ulimit -S -t 3600
>
> which then applied to all child build scripts.  I was about to add a new
> feature to bob to support arbitrary precmds, but on reflection felt that
> this would actually be better as a native pkgsrc feature so that it
> works everywhere and not just in bulk builds.
>
> My initial draft is to add a RUN_PRECMD hook to ${RUN}, so that I can
> set e.g.
>
>    RUN_PRECMD=  ulimit -S -t 3600
>
> in mk.conf and everything should Just Work.
>
>    --- a/mk/misc/common.mk
>    +++ b/mk/misc/common.mk
>    @@ -10,0 +11,5 @@
>    +# RUN_PRECMD
>    +#      Commands that are placed at the end of the expanded ${RUN} and before
>    +#      the main body of commands.  This is a helpful place to insert e.g.
>    +#      'ulimit' to avoid hung processes causing builds to run indefinitely.
>    +#
>    @@ -30,0 +36 @@ PKG_DEBUG_LEVEL?=       0
>    +RUN_PRECMD?=           :
>    @@ -44 +50 @@ _PKG_DEBUG_SCRIPT=        ${SH} -x
>    -RUN=                   ${_PKG_SILENT}${_PKG_DEBUG} set -e;
>    +RUN=                   ${_PKG_SILENT}${_PKG_DEBUG} set -e; ${RUN_PRECMD};
>
> Any better ideas?  Would a more specific LIMIT_RESOURCES feature that is
> analogous to UNLIMIT_RESOURCES be more appropriate, or just too complex?

I like the flexibility of RUN_PRECMD. Maybe just expand the comment a
little to give a casual reader more guidance?

+#      the main body of commands.  This is a helpful place to insert e.g.
+#      'ulimit -S -t 3600' to abort after 3600 seconds to avoid hung processes
+#      causing builds to run indefinitely.

David