view tests/test-topic-issue6500.t @ 6197:f1d4c9eddcc1

tests: use desc() revset instead of node hashes in test-evolve-continue.t

Using desc() makes the tests less fragile if new steps are inserted, since the
test output can be more easily regenerated rather than having to go and find
all of the commit hashes in the input to replace them.

KeyError: b'topic' on history-rewriting commands (issue6500)
https://bz.mercurial-scm.org/show_bug.cgi?id=6500

  $ . $TESTDIR/testlib/common.sh

Making sure we're not caching .topic() results for memctx or anything else that's not stored on-disk

  $ hg init issue6500-caching-memctx
  $ cd issue6500-caching-memctx

  $ cat >> $HGRCPATH << EOF
  > [extensions]
  > evolve =
  > topic =
  > EOF

for this test we need 2 changesets with amend_source, one with topic and one without

  $ hg topics foo
  marked working directory as topic: foo
  $ echo apple > a
  $ hg ci -qAm 'apple'
  $ echo apricot > a
  $ hg ci --amend -m 'apricot'

not using `hg topics --clear -r .` here because that would remove amend_source, see _changetopics()

  $ hg topics --clear
  $ hg ci --amend -m 'no foo apricot'

  $ hg log --hidden -r 1+2 -T '{rev}: {join(extras, " ")}\n'
  1: amend_source=* branch=default topic=foo (glob)
  2: amend_source=* branch=default (glob)

creating and handling 2 memctx instances (based on 1 and then 2) should work

  $ hg touch --hidden -r 1+2 --duplicate
  switching to topic foo

make sure extras stay the same

  $ hg log --hidden -r 3+4 -T '{rev}: {join(extras, " ")}\n'
  3: __touch-noise__=* amend_source=* branch=default topic=foo (glob)
  4: __touch-noise__=* amend_source=* branch=default (glob)