On Thu, Mar 12, 2026 at 08:13:08AM +0100, Rene Kita wrote:
On Wed, Mar 11, 2026 at 06:57:11PM +0800, Kevin J. McCarthy wrote:
I squished the 3rd and 4th commit together.

I also added the cases from the attachment menu (forwarding and replying)
that generate a tempfile and pass it in.

Quickly tested on OpenBSD 7.8, seems to work.

I'm not convinced this is worth the effort and extra code, but I'm OK
with it if Kevin wants to have it.

It was just swapping the mktemp calls in three places, so no biggy. I just forgot about those cases because I (personally) never use them.

That being said, without that patch mutt will respect TMPDIR when
creating the draft file. With the patches applied TMPDIR is ignored.
This might break existing workflows.

LOL I just replied to the gitlab comment by Raf. He had the same question.

I'll just paste my reply again:

It seemed to me that if mutt's $tmpdir already defaulted to $TMPDIR, and we wanted to distinguish $tmpdraftdir as a more persistent temporary place, it made sense to not also set it to the env var by default when set.

So my thinking was defaulting it to the non-volatile "/var/tmp", but the user can just as easily set it to something else in their muttrc.

Also as /var/tmp is not defined by POSIX yet, are we really sure it will
exist on all systems?

I thought, although not strictly defined, it was pretty well established for a long time. But if this really is an issue we could do some tests
in configure.ac.  What other directories would be equivalent?

--
Kevin J. McCarthy
GPG Fingerprint: 8975 A9B3 3AA3 7910 385C  5308 ADEF 7684 8031 6BDA

Attachment: signature.asc
Description: PGP signature

Reply via email to