push: rework the computation of fallbackheads to be correct
The previous computation tried to be smart but ended up being wrong. This was
caught by phase movement test while reworking the phase discovery logic to be
faster.
The previous logic was failing to catch case where the pushed set was not based
on a common heads (i.e. when the discovery seemed to have "over discovered"
content, outside the pushed set)
In the following graph, `e` is a common head and we `hg push -r f`. We need to
detect `c` as a fallback heads and we previous failed to do so::
e
|
d f
|/
c
|
b
|
a
The performance impact of the change seems minimal. On the most impacted
repository at hand (mozilla-try), the slowdown seems mostly mixed in the
overall noise `hg push` but seems to be in the hundred of milliseconds order of
magnitude. When using rust, we seems to be a bit faster, probably because we
leverage more accelaratd internals.
I added a couple of performance related common for further investigation later
on.
Setup:
$ cat > eval.py <<EOF
> import filecmp
> from mercurial import commands, context, pycompat, registrar
> cmdtable = {}
> command = registrar.command(cmdtable)
> @command(b'eval', [], b'hg eval CMD')
> def eval_(ui, repo, *cmds, **opts):
> cmd = b" ".join(cmds)
> res = pycompat.bytestr(eval(cmd, globals(), locals()))
> ui.warn(b"%s" % res)
> EOF
$ echo "[extensions]" >> $HGRCPATH
$ echo "eval=`pwd`/eval.py" >> $HGRCPATH
Arbitraryfilectx.cmp does not follow symlinks:
$ mkdir case1
$ cd case1
$ hg init
#if symlink
$ printf "A" > real_A
$ printf "foo" > A
$ printf "foo" > B
$ ln -s A sym_A
$ hg add .
adding A
adding B
adding real_A
adding sym_A
$ hg commit -m "base"
#else
$ hg import -q --bypass - <<EOF
> # HG changeset patch
> # User test
> # Date 0 0
> base
>
> diff --git a/A b/A
> new file mode 100644
> --- /dev/null
> +++ b/A
> @@ -0,0 +1,1 @@
> +foo
> \ No newline at end of file
> diff --git a/B b/B
> new file mode 100644
> --- /dev/null
> +++ b/B
> @@ -0,0 +1,1 @@
> +foo
> \ No newline at end of file
> diff --git a/real_A b/real_A
> new file mode 100644
> --- /dev/null
> +++ b/real_A
> @@ -0,0 +1,1 @@
> +A
> \ No newline at end of file
> diff --git a/sym_A b/sym_A
> new file mode 120000
> --- /dev/null
> +++ b/sym_A
> @@ -0,0 +1,1 @@
> +A
> \ No newline at end of file
> EOF
$ hg up -q
#endif
These files are different and should return True (different):
(Note that filecmp.cmp's return semantics are inverted from ours, so we invert
for simplicity):
$ hg eval "context.arbitraryfilectx(b'A', repo).cmp(repo[None][b'real_A'])"
True (no-eol)
$ hg eval "not filecmp.cmp(b'A', b'real_A')"
True (no-eol)
These files are identical and should return False (same):
$ hg eval "context.arbitraryfilectx(b'A', repo).cmp(repo[None][b'A'])"
False (no-eol)
$ hg eval "context.arbitraryfilectx(b'A', repo).cmp(repo[None][b'B'])"
False (no-eol)
$ hg eval "not filecmp.cmp(b'A', b'B')"
False (no-eol)
This comparison should also return False, since A and sym_A are substantially
the same in the eyes of ``filectx.cmp``, which looks at data only.
$ hg eval "context.arbitraryfilectx(b'real_A', repo).cmp(repo[None][b'sym_A'])"
False (no-eol)
A naive use of filecmp on those two would wrongly return True, since it follows
the symlink to "A", which has different contents.
#if symlink
$ hg eval "not filecmp.cmp(b'real_A', b'sym_A')"
True (no-eol)
#else
$ hg eval "not filecmp.cmp(b'real_A', b'sym_A')"
False (no-eol)
#endif