(LTH-166) Check if waitpid was merely interrupted

[aduitsis]

A signal may interrupt a system call such as waitpid, in which case it would return -1 but with errno set to EINTR. The typical solution is to run waitpid inside a while loop and continue looping while waitpid returns -1 with the errno variable is set to EINTR. This patch does exactly that.

This patch fixes this particularly nasty bug: https://tickets.puppetlabs.com/browse/FACT-2019, but one could easily imagine other strange bugs (or rather, Heisenbugs!!!) occurring as a result.

[puppetlabs-jenkins]

Can one of the admins verify this patch?

[GabrielNagy]

Hi @aduitsis,

Thanks for your contribution! I've changed the project of the ticket you opened to Leatherman, since the fix is being done here.

Two things:

• please retarget your PR on the 1.4.x branch, so this gets in all our release streams
• please amend your commit title to include the ticket name (i.e. (LTH-166) Check if waitpid was merely interrupted)

Thank you!

[puppetcla]

CLA signed by all contributors.

[aduitsis]

Hi @aduitsis,

Thanks for your contribution! I've changed the project of the ticket you opened to Leatherman, since the fix is being done here.

Two things:

• please retarget your PR on the 1.4.x branch, so this gets in all our release streams
• please amend your commit title to include the ticket name (i.e. (LTH-166) Check if waitpid was merely interrupted)

Thank you!

Hello,

Changed the name, retargeted to branch 1.4.x. Note, I took it from the master and sent it to 1.4.x, is that okay?

Thanks!

[puppetcla]

CLA signed by all contributors.

[allanjude]

I can confirm this issue with leatherman was breaking facter and puppet on FreeBSD with ruby 2.6. Applying this patch resolves the issue.

More detail and debug output in our issue: https://bugs.freebsd.org/bugzilla/show_bug.cgi?id=240566

[GabrielNagy]

looks pretty straightforward to me 👍