[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index][Old Index]
Re: Timeout on configure stage
* On 2020-11-08 at 02:13 GMT, Joerg Sonnenberger wrote:
> On Sun, Nov 08, 2020 at 02:17:32AM +0100, Sijmen Mulder wrote:
> > Still trying to get a full bulk build going on macOS/aarch64, I hit
> > upon a hang in textproc/ja-groff. I found that one of its configure
> > tests ('checking for working mmap...') hangs on a read() on this
> > configuration.
> > Putting the specifics of that issue aside, it seems problematic to
> > me that this locks up the entire bulk build. I suggest implementing
> > a timeout on the configure stage of say, 10 minutes, to rule out
> > this mode of breakage. Would that be ok?
> I've been recommending bulk builds to use an appropiate CPU time limit
> (ulimit -t) for years. What is appropiate depends a bit on the CPU, of
> course, but 60min is a good starting point. It doesn't cover all cases
> and the problematic cases are those where no good out-of-the-box
> solution in Unix exists.
In addition to this (I also add "ulimit -t 3600" to my pbulk.conf)
there are some packages that simply hang without using CPU, and
there's no reasonable way to detect this.
For these, which include textproc/ja-groff, I have this section in my
mk.conf that skips such packages entirely:
You will probably want similar. Some of these packages may have since
been fixed, though, so we should verify the full list at some point.
Jonathan Perkin - Joyent, Inc. - www.joyent.com
Main Index |
Thread Index |