1 patch for repository darcs.net:/opt/darcs/screened:

patch 152f3531540735a10947d2a1db5043117a26fd7b
Author: Ganesh Sittampalam <ganesh@earth.li>
Date:   Wed Jun 21 08:53:03 BST 2023
  * use --ignore-times in rebase-amend test

  This test fails on my machine after

  patch b0ec10c78ac19423c896352ed7fac14b59f60229
  * enable use of the index in the test suite by default

  But I'm not sure why some tests require --ignore-times
  and others don't.

Attachments

• patch-preview.txt
• use-__ignore_times-in-rebase_amend-test.dpatch

This appears to fix http://bugs.darcs.net/issue2705 for me.

The correct fix for this and some other failing tests is to not 
include stderr in the log of the output, i.e. remove trailing " 
2>&1" when we capture the output of a darcs command. Doing this for 
the affected tests reduces the number of failing tests to one, 
namely tests/add.sh; which is a special case, since we explicitly 
test that nothing is printed to stderr; this particular part of the 
test script should be removed or re-written using some other method 
to make sure the command did what it should.

The reason these tests do not fail for screened is

patch dc1f597e0c48fec5a415d6e5f0bcd342f5717331
Author: Ben Franksen <ben.franksen@online.de>
Date:   Sun Mar 12 10:55:16 CET 2023
  * progress reporting: fix for non-native Windows terminals and 
simplify

which among other changes turns off progress reporting completely 
when stderr is redirected.

Correction: as stated in the patch comment, progress reporting now 
goes to stdout and only if that is a terminal device.

On Linux this patch fixes all test failures in screened. If it works 
on your new Windows machine, too, then we should fast-track 
accepting Patch2291.

I think my two different test failure reports have got mixed up.

http://bugs.darcs.net/issue2705, which this patch tries to fix,
is specifically to do with intermittent index corruption, not
extra lines in display output.

I've double-checked, and pulling patch2291 on top of reviewed,
along with a single dependent patch ("update outdated dependencies"),
doesn't fix the problem.

Specifically, 

cabal run test:darcs-test -- -t rebase-amend

fails intermittently in the final 'darcs check'.

I am running the CI to see if this can reproduce the failure you see.

Thanks for reminding me I should check screened, where the test does
in fact work.

I did some manual bisection and it appears to be fixed by this.
I'm a bit confused about the reasons though!

    patch 9af249013964d18a07dcf371a258b58e771e27d2
    Author: Ben Franksen <ben.franksen@online.de>
    Date:   Sun Feb 26 21:43:17 GMT 2023
      * rebase: take applyToWorking out of doSuspend

What 9af2490139 does is to "normalize" how we treat changes to the working 
tree when we suspend patches: the standard way is to first finalize the 
"internal" repo changes (i.e. update inventory and pristine tree), then 
afterwards apply changes to the working tree. Whereas before this patch, we 
changed working inside doSuspend, and finalized afterwards.

Note that finalizeRepositoryChanges also updates the index, which does look 
at the working tree.

ATM these are only data points, I don't have a complete explanation yet.

I've followed up on the details of the failure in 
http://bugs.darcs.net/issue2705

Marking this one as rejected since we have a fix in screened.