bad state after failure / nonworking mail server

[John Marino]

Hi guys,

My tinderbox/pkgsrc runs along fine, building package after package
until one has a problem, and then it doesn't recover.  The build stops,
but the duration clock continues, and all the null mounts stay mounted.
 I had to clean all this up manually.

So I went to see if I could figure out what's wrong.  Here's what I
think is going on:

Given: p5-Net is installed, but there's no working mail server here.

So I think the script gets back to lib/portbuild and executes line 474
which is "mark_failed ... ".
Then it tries to send an email, which fails.
I think the tc_command handles tells the script to bail out, so the next
two lines are never executed:

> execute_hook "postPortBuild" "PACKAGE_NAME=${pkgname} BUILD=${build} JAIL=${jail} (MORE)
> cleanup ${chroot} ${noclean} ${error} "${cleandirs}" ${pkgname} ${build} ${nullfs}

Then I commented out the line before (${tc} sendBuildErrorMail -d
${portdir} -b ${build} -p ${pkgname} -x ${lext}) and confirmed that
tinderbox cleaned up after failed packages after that change.

Can somebody confirm the (mis)handling of email errors is leaving
Tinderbox in an unclean state?

If this is confirmed, I think the handling should be changed, or at
least have a configuration option to NOT try to send email.

Regards,
John