Ian Jackson writes ("Re: Bug#1094634: Info received (autopkgtest-virt-*:
doesn't shutdown on EOF [and 1 more messages])"):
> If this is in fact getting EAGAIN it probably indicates that something
> was setting stdin to nonblocking mode? The workaround is quite wrong.
I conjecture:
* Something, somewhere, was setting some autopkgtest-virt-*
program's stdin, to nonblocking mode. (Code in the autopkgtest
tree itself doesn't seem to do that.)
* The checks for EAGAIN, and the loop on '', were put into
VirtSubproc as a workaround for this.
I propose the following, fairly conservative, change:
* Look for whether stdin is nonblocking, at startup.
* If it is, print a scary warning to autopkgtest's stderr,
saying that this is a bug in whatever called us, and that this will
not be supported in the future.
* And, in that case, enable the existing workarounds.
* Otherwise, go back to the previous behaviour, where errors and EOF
on stdin cause an immediate error and teartodnw.
I intend to make this into an MR on salsa.
Ian.
--
Ian Jackson <ijack...@chiark.greenend.org.uk> These opinions are my own.
Pronouns: they/he. If I emailed you from @fyvzl.net or @evade.org.uk,
that is a private address which bypasses my fierce spamfilter.
