This way export_commit is much simpler (already quite complex), and it's
easier to modify the logic.
No functional changes.
Signed-off-by: Felipe Contreras <felipe.contreras@gmail.com>
There's no need to call repo[revnode] when repo[rev] works perfectly
fine.
And since we have the context already we can just do ctx.hex() instead
of hexlifying ourselves.
Signed-off-by: Felipe Contreras <felipe.contreras@gmail.com>
The `git init` command can create the directory, and HEAD doesn't need
to be specified in `git checkout` (it's the default).
Signed-off-by: Felipe Contreras <felipe.contreras@gmail.com>
The length should be exactly the same as the data, for example if the
data is "hello" only 5 characters should be written on the stream. Thus
it should always be `len(data)`, not `len(data)+1` as it currently is in
some places.
Since the first commit of hg2git.py there was a wtf comment, presumably
Rocco was confused about this common discrepancy.
We can shuffle the logic around by adding '\n' to the data, and removing
+1 to the length.
Also, the data should be written without a newline (wr_no_nl).
Signed-off-by: Felipe Contreras <felipe.contreras@gmail.com>
In `process_unicode_escape_sequences()`, any backslash escape sequences
in the original string are escaped upon the first
`.encode('unicode-escape')` and therefore round-trip the sequence of
`.encode('unicode-escape').decode('unicode-escape')`.
That is not what we want - we want these sequences to be passed-through
the `.encode` unchanged, so that they will be converted to the
character they represent upon `.decode()`.
This patch changes the `.encode()` step to pass through any ascii
characters unchanged, only escaping non-ascii characters. This ensures
any existing backslash escape sequences will be interpreted as the
character they represent upon `.decode()`.
Since Python 3.7 the re module warns for syntax which could, in the
future, be misparsed as a nested set. Avoid this by escaping the
literal `[` we search for in the regexp.
Reported by Monte Davidoff @mndavidoff
Closes#269.
If fast-export was asked to export a Mercurial branch to Git and a
branch of the same name already existed in the Git repo but it was not
created by fast export, fast-export would crash while trying to format
an error message claiming that the destination branch was modified
behind its back.
This patch extends fast-export to detect the situation above and give
a proper error message which hopefully is less confusing to the user.
Credits for discovering the original crash goes to Shun-ichi Goto
<gotoh@taiyo.co.jp>.
Closes: #269.
Try to make it clear that sloppy, throw it over the fence, patches
won't be accepted without revision and try to make sure a potential
contributor sees the warning while creating a pull request.
Keys and values in the state cache are byte strings, therefore a
lookup of 'tip' will always fail. The failure makes the conversion
start over from the beginning, but as fast-export is deterministic the
results are the same, just very inefficient. The bug has existed since
the port to Python 3.
This patch switches the 'tip' lookup to use a byte string which should
make incremental conversions restart at the last converted commit. As
'x' == b'x' in Python 2, this should be a backwards compatible change.
Bug reported and fix suggested by Tomas Kolda.
Fixes#258.
