On Wed, Apr 08, 2009 at 11:50:23AM -0700, Danek Duvall wrote:
> I don't know that we'll run into this much with Bart's recent changes, but
> I'm sure we'll bulk things up again at some point, and I'm sure people will
> still try to run pkg in ridiculously low memory environments, so it'll
> still be useful.
> 
>     http://cr.opensolaris.org/~dduvall/pkg-memoryerror/

Unless I've misunderstood, it looks like this fix only catches a certain
subset of the memory errors that we care about.  2297 shows clients
getting an OSError with ENOMEM as a possible case.  It doesn't look like
this fix addresses the situation where a system call fails becuase of
low memory.  FWIW, posix_spawn(3C) can still return ENOMEM, as any of
the errors from fork(2) or exec(2) are fair game.

If you go to the trouble to catch an ENOMEM, it might also be worthwhile
to try to catch an ENOSPC and print a similar message for the filesystem
that ran out of space.  IIRC, file-based environment errors may contain
the path to the file.  I understand if you don't want to fix that in
this wad, though.

-j
_______________________________________________
pkg-discuss mailing list
[email protected]
http://mail.opensolaris.org/mailman/listinfo/pkg-discuss

Reply via email to