fish 3.0.1?

[zanchey]

I'm asking for opinions on whether to put out a 3.0.1 to fix showstopping bugs following 3.0.0, and what to include. I would be aiming to get it out this month. Otherwise they will have to wait for 3.1.0, which I've got planned for March or so.

Candidates include:

• Exec inside a function does not work #5449 (regression)
• Fish 3.0 does not run under Cygwin #5426 (build failure on Cygwin)
• Fish goes 100% CPU when I go into "less" mode in history #5438 (regression)
• read --silent is broken in Fish 3.0 #5519 (regression, unfixed as of this writing)
• fish 3.0 SIGSEGV under FreeBSD 12 #5453 (possibly - memory corruption, hit in the tests, could conceivably be hit in production)
• Error: Wrong color in test at index 3 in text (expected 0x104, actual 0x1): cd test/fish_highlight_test #5458 (possibly - failure on OpenIndiana)
• GLIBCXX_ASSERTIONS reveals problem in string match subcommand #5479 (memory corruption segfault, hit in the tests, could conceivably be hit in production)

[mqudsi]

I think the OpenIndiana folks are doing a fine job incorporating patches and updates post 3.0, so #5458 isn't very important. The others all sound worth including.

[cdluminate]

I (Debian developer maintaining fish) second the request for the 3.0.1 release. We are reaching the freezing stage of the next stable release (aka. Debian Buster) [1]. And the timing of 3.1.0 release sounds too late for Debian Buster. So I think releasing 3.0.1 would make great sense.

[1] https://release.debian.org/

[faho]

I'm all for doing a bugfix release!

The question is always how big we want it to be?

I would say anything from https://github.com/fish-shell/fish-shell/issues?utf8=%E2%9C%93&q=is%3Aissue+is%3Aclosed+milestone%3A%22fish+3.1.0%22+ is a candidate here.

Some things I think should be included:

• /etc/paths.d entries are split when containing spaces #5481 is a regression and an easy fix

• Many getpid calls wasting CPU resources  #5447, at least a7998c4 is a performance regression fixed by removing an assert.

And some maybes:

• Git filename completion only work with starting pattern #5476, given that it's been reported twice. It's more of a matter of taste, though.

• return inside while does not return correct status #5513 would be nice to include, if we can get it fixed in a timely manner

• Fish subshell changes PATH order #5456 would be nice to have, if we can come up with an answer for it. It's slightly larger, though.

[floam]

Yes, I think we should, there are a few real stinkers that would be nice to get out of the way so we can focus on having time to do a nice 3.1.

[mqudsi]

I personally view #5513 as worthy of a 3.0.1 all on its own, but maybe that's just me.

[ridiculousfish]

I've got two weeks free, I can definitely do this.

[ridiculousfish]

master has diverged too far so 3.0.1 should be based on 3.0.0. I've added an Integration_3.0.1 branch to origin, let's fix the 3.0.1 issues in master, and cherry-pick into Integration_3.0.1.

[ridiculousfish]

Milestone is complete, let's do it!

[zanchey]

Awesome - well done. I should have time to roll the release in the next week or so (perhaps even tonight).

[ridiculousfish]

Thanks everyone for the hard and timely work here!