A dumber approach to reverting #7995 on 3.8 branch

[Mikolaj]

A dumber (commit by commit) approach to reverting #7995 on 3.8 branch. This is made painful by the fact that we got sloppy and didn't rebase a subsequent PR and so the overall merge got complex instead of fast-forward. Also, some other PRs (at least one, but probably more) actually depend on this PR.

Manual hacks notes:

• I've left the definition of rawSystemProcAction and fromCreatePipe in the revert of Rework subprocess helpers, because they were causing a conflict and indeed they are used in some subsequent PRs
• I've re-added logCommand (from the same reverted commits as above) that is used in newer PRs

[jneira]

Thanks for trying the revert, ci is failing at UNEXPECTED FAIL: PackageTests/NewBuild/CmdRun/Terminate/cabal.test.hs, like mine one: i had to revert #8081 with a796fa9

[Mikolaj]

Let's hope your version works. I remember CI was a flaky with that test, but it's better than failing every time. I'm trying another hack, but at this point it's failing outright. :)

[Mikolaj]

This really failed hard with "*** Exception: unexpected: done sleeping", so changing the test harness doesn't seem to be sufficient. The cleanup probably fixed the real code that is used in the test. I will try your idea and revert the change of the test #8081 that adjusts it to the revised code (that we are reverting here).

[Mikolaj]

The same failure with #8081 reverted, so I'm just going to mark the test as known broken. It was flaky before.

[Mikolaj]

If CI passes, I'm going to tag this to get gitlab to build binaries so that @jneira can easily test if this works around #8208.

[robx]

Here's the list of commits that you'd want to revert. It's not a good idea to keep #7921 if reverting #7995, that's very likely to break things.

2521ee9d4 - Fix race condition in PackageTests/NewBuild/CmdRun/Terminate (4 months ago) <Robert Vollmert>
aa4f0ad3c - Add changelog (4 months ago) <Robert Vollmert>
d2172d896 - Add test for terminating "cabal run" on unix (4 months ago) <Robert Vollmert>
a03b4a79e - Handle SIGTERM by throwing an asynchronous exception (4 months ago) <Robert Vollmert>
0284bfbec - Use Process.withCreateProcess to ensure child cleanup (4 months ago) <Robert Vollmert>
67092dab6 - Use maybeExit in more places (4 months ago) <Robert Vollmert>
91fa33bd5 - Add changelog (5 months ago) <Robert Vollmert>
97f2c7b38 - Remove newly deprecated functions (5 months ago) <Robert Vollmert>
d5ae20cf9 - Deprecate obsolete functions (5 months ago) <Robert Vollmert>
9d33338c1 - manpage: use rawSystemProcAction (5 months ago) <Robert Vollmert>
3aa45a2d7 - LibV09: use rawSystemProcAction (5 months ago) <Robert Vollmert>
200478ef9 - SetupWrapper: replace runProcess' by rawSystemProc (5 months ago) <Robert Vollmert>
1969327c8 - Set delegate_ctlc for rawSystemStdInOut, too (5 months ago) <Robert Vollmert>
f972102ef - Default delegate_ctlc to True (5 months ago) <Robert Vollmert>
a5986e37f - Rework subprocess helpers (5 months ago) <Robert Vollmert>
eefe45d9f - Drop unused module Distribution.Client.Compat.Process (5 months ago) <Robert Vollmert>
592a48e33 - Disambiguate qualified imports (5 months ago) <Robert Vollmert>

[Mikolaj]

OK, let's do that in yet another PR.

[Mikolaj]

Tried it, at least not conflicts this time...

[Mikolaj]

Done on #8321.

[Mikolaj]

CI passed fine, but closing in favour of #8321.