| Age | Commit message (Collapse) | Author |
|---|
|
The change made to RC1 for fixing diff uploading in some cases actually
broke things. The content length was no longer correct, if we even could
encode to UTF-8. We now upload the raw bytes once again.
|
|
If reading a diff failed and we received a one-line error message,
post-review would crash. It assumed a second line always existed. Now
we properly abort with debug information.
Patch by Laurent Nicolas.
Reviewed at http://reviews.reviewboard.org/r/1457/
|
|
The new flexible HTTP error handling code had an uninitialized variable. This
checks for it and makes sure not to break.
Patch by Laurent Nicolas.
Reviewed at http://reviews.reviewboard.org/r/1427/
|
|
First, using --revision-range with 0 wouldn't work, as the path given
wouldn't have existed in revision 0. The repository path is now assumed
if using revision 0, in order to get a diff of all the files.
Second, sometimes an svn diff would have excess tabs in the diff file headers,
which we wouldn't catch correctly. We now make sure to only split the headers
into two groups, regardless of the number of tabs.
Patch by Nathan Heijermans.
Reviewed at http://reviews.reviewboard.org/r/1429/
|
|
The new default changeset support was failing to set a default changeset
description on newer versions of P4, which we would later need to parse
in order to get the file listing. We now go through the same code path we went
through with older versions of P4 in order to generate a default description
in order to get those generated files.
Patch by Laurent Nicolas.
Reviewed at http://reviews.reviewboard.org/r/1418/
|
|
The ClearCase code was calling a command and expecting the result split
across two variables, but didn't pass split_lines=True. Now it does!
Patch by Bryan Halter
Fixes bug #1423
|
|
|
|
|
|
|
|
We filtered out the username from a path when running 'hg svn info', but
not when reading from .hgrc. We now apply the same filtering logic. This
allows post-review to match a repository path with a username in it against
a repository path on a Review Board server.
Patch by Tom Saeger.
Reviewed at http://reviews.reviewboard.org/r/1390/
|
|
This change allows --diff-filename to be used with --repository-url. This
can be used to submit a diff outside of a working checkout.
Patch by Luke Robison.
Reviewed at http://reviews.reviewboard.org/r/1393/
|
|
Patch by Stephen Gallagher.
Reviewed at http://reviews.reviewboard.org/r/1384/
|
|
Patch from Petr Novák. Current post-review's GitClient calls "git remote show"
to determine URL of upstream repository and parses its output. This patch
replaces this with fetching the URL from local configuration, which spares a few
moments of time since "git remote show" has to communicate with the remote in
order to determine list of remote branches (which is subsequently discarded by
the script).
Testing done:
Tested locally, seems to work fine. (I used it to submit this review request).
Reviewed at http://reviews.reviewboard.org/r/1377/
|
|
Patch from Thilo-Alexander Ginkel. Previous post-review versions did not
correctly generate the Content-Length header sent along with the
api/json/reviewrequests/<id>/diff/new/ request as the header's value was
calculated based upon the length of the request body in characters instead of
bytes. This patch should fix that issue by using the UTF-8 encoding of the
body for the length calculation.
Reviewed at http://reviews.reviewboard.org/r/1298
|
|
|
|
Patch from Laurent Nicolas. post-review does not process a *pending* Perforce
changeset in v0.2 beta3 (Note that it is a regression from v0.2 beta2).
Also 2 changes that are not strictly related:
1) allow *pending* to be on the second line in addition to the first line
2) append the list of bugs to existing list if it exists already rather than
overwritting it.
Note the "view diffs" view does not show a white space change on line 1586. I
fixed the indentation, otherwise the for loop was ending after the first line.
Testing done:
Tested post-review against our perforce repository, using two *pending*
changeset. Tried combination of with and without -r to find existing review
based on changeset number or review id. Added bug list to check append feature.
Checked for different values of --bugs-closed. After each invocation, checked
the GUI for a correct list, and then submitted the review before the next try:
1) No option
Bugs: empty
2) --bugs-closed=123
Bugs: 123
3) --bugs-closed=456
Bugs: 123, 456
4) --bugs-closed=456 (repeat)
Bugs: 123, 456
5) --bugs-closed=12
Bugs: 12, 123, 456
6) --bugs-closed=1234
Bugs: 12, 123, 456, 1234
7) --bugs-closed="1234, 12345, 1,"
Bugs: 1, 12, 123, 456, 1234, 12345
8) --bugs-closed="1,, 234, 12345, 1, ,,"
Bugs: 1, 123, 234, 456, 1234, 12345
10) --bugs-closed=" ,, ,,1,, 234, 12345, 1, ,, , , ,,,, ,,,,,"
Bugs: 1, 12, 123, 234, 456, 1234, 12345
Checked the database: Bugs field: 1234,12,456,1,123,234,12345
Fixes bug 1345
|
|
post-review's HTTP GET/POST code always assumed that any Review Board API
error code would be in an HTTP 200, and that anything else was an unknown
failure. It's now more flexible and understands that any HTTP status code
could have a JSON payload in it, and processes them accordingly.
This introduces some new unit tests for testing the error parsing code
paths.
Reviewed at http://reviews.reviewboard.org/r/1375/
|
|
This adds support for passing diff content on STDIN when using
--diff-filename=-. When specified, any input that is required (such as
authentication information) will terminate and inform the user they must pass
the appropriate information on the command line.
|
|
This new parameter allows the caller to override the diff being uploaded to
the Review Board server. This diff will be used instead of generating a diff
manually.
Patch by Chris Clark
Reviewed at http://reviews.reviewboard.org/r/1197/
|
|
We were using lstrip to remove the prefix 'refs/heads/', but that strips
all characters in the string, not the string itself. We now properly strip
this prefix.
Patch by Dan Savilonis
Reviewed at http://reviews.reviewboard.org/r/1360/
|
|
post-review makes an assumption that the 'master' branch in a Git repository
represents SVN's 'trunk' when using git-svn. Now it uses the calculated
parent branch, which may be 'master' or may be some other branch, depending
on the repository setup.
Patch by Mike Crute
Reviewed at http://reviews.reviewboard.org/r/1307/
|
|
This implements support for --guess-description and --guess-summary with
revision ranges in Mercurial. The commit messages are concatenated to form
the description, and the first line of the first commit message forms the
summary.
Patch by Stefan Ring
Reviewed at http://reviews.reviewboard.org/r/1333/
|
|
A recent change to post-review broke --guess-description with Git when not
specifically using a parent branch. We now default to grabbing all logs from
the merge base (for example, "master") and using that as the description. This
covers the same range as the diff itself.
|
|
New versions of Perforce provide more explicit information when moving
files. There are two new types: move/add and move/delete. These more or
less map (for our purposes) to just add and delete. We should now handle
them.
Fixes bug #1417
|
|
Some Perforce setups have DNS aliases for a Perforce server. We currently
fetch the server aliases, but weren't doing anything with them. Now we support
scanning a list of aliases against the known list of repositories on Review
Board in order to find the right repository URL to use. This happens
automatically without any additional configuration.
Patch by Ravi Kondamuru
Reviewed at http://reviews.reviewboard.org/r/1235/
|
|
A recent patch introduced a refernece to HttpDigestAuthHandler, but this
should have been HTTPDigestAuthHandler.
Patch by Jason Felice
Reviewed at http://reviews.reviewboard.org/r/1290/
|
|
Patch from Ryan Oblak to add support for HTTP Digest Authentication to
post-review. It was really trivial given the work already added in for HTTP
Basic Authentication.
Testing done: This has been tested on my Review Board 1.1 alpha 1 installation
with Digest, Basic, and no authentication.
Reviewed at http://reviews.reviewboard.org/r/1265/
|
|
We need GitClient.head_ref to be defined for diffing, but this was only
defined in the pure Git case. We now set this before determining the mode
for GitClient.
Fixes bug #1400
|
|
We were requiring the simplejson dependency on Python 2.6, despite the
fact that Python 2.6 includes json built-in. We no longer depend on simplejson
on this version.
Fixes bug #1383
|
|
When adding/changing binary files against a git-svn repository, the binary
files would be filtered out of the diff. We now represent binary files
properly by faking the information in a diff.
Patch by Anthony Cruz.
Reviewed at http://reviews.reviewboard.org/r/1194/
|
|
Running post-review --diff-only may throw exceptions when encountering
errors, instead of gracefully handlin them. A simple case of this is to
run with --diff-only without having a post-review login cookie, and instead
of a login prompt, post-review will crash with a KeyError.
This change fixes this reporting by not special-casing the --diff-only
parameter in the error reporting. We were only handling errors if that
option was not set.
Patch by Eric Huss.
Reviewed at http://reviews.reviewboard.org/r/1185/
|
|
This adds improved logic for determining the parent branch used when generating
a diff. It will attempt to detect the merge base of the current head, falling
back to defaults if necessary. A tracking branch can be manually specified
too.
The new code attempts the following:
1) Tries to detect the tracking branch for the current HEAD, if possible.
2) If a tracking branch doesn't exist, or is not remote, it will fall back
on 'origin/master'.
3) If the --tracking-branch parameter is specified, this branch will be used
instead of 'origin/master'.
It also provides unit tests for post-review. Right now, these test only
GitClient (in pure Git mode), but new tests can be added down the line for
the rest of post-review.
Patch by Dan Savilonis
Reviewed at http://reviews.reviewboard.org/r/1144/
|
|
|
|
|
|
If a .reviewboardrc file had a syntax error, we would silently ignore it,
which made it hard to debug problems. We now fail on syntax errors and
list the error information, including the filename, line number and offset on
the line.
Patch by Eric Huss
|
|
The fix for bug #585 broke the binary file support. This change brings it
back.
This also fixes the Perforce server versioning check.
Patch by Eric Huss
Reviewed at http://reviews.review-board.org/r/1154/
|
|
Some people have source files that don't end in a newline. In this case, it's
possible for us to create a diff file where the end of one file is joined on the
same line as the header for the next file. This change adds an extra check to
_do_diff to make sure that we don't make broken diffs when people have broken
files.
Fixes bug 1340.
|
|
When creating an SVN diff for a remote repository using --revision-range and
passing in specific filenames, post-review would only get the first file as the
diff. This is because it was using the arg as the base path. This change makes
it treat the arg as a file list if it's more than one item long.
Fixes bug 1162
|
|
Patch from Flavio Castelli. With this patch post-review will support the
revision-range option following the syntax explained at
http://www.review-board.org/docs/manual/dev/users/tools/post-review/#posting-committed-code
Testing done:
I have tested this new feature locally and it works fine.
I have successfully tested the following commands:
post-review --revision-range=rev1:rev2 --guess-summary --guess-description
post-review --revision-range=rev1:rev2 --guess-description
post-review --revision-range=rev1:rev2
post-review --revision-range=rev --guess-summary --guess-description
post-review --revision-range=rev
Reviewed at http://reviews.review-board.org/r/1049/
|
|
When setting a field (such as target reviewers) from post-review, we were
saving the draft, always. This is bad enough, but we were doing this before
even updating the diff. This would result is a published draft and an extra
change description entry. We no longer save the draft, unless the user
uses --publish.
Fixes bug #811
|
|
We were hard-coding the path to the user's Application Data directory, which
changed in Windows Vista. We now use the APPDATA environment variable, which
is guaranteed to point to the correct directory.
Fixes bug #1279
|
|
Patch from Matthew Woehlke. This fixes post-review to work on older p4d (which I
am stuck with at $DAYJOB); without this change post-review can only be used on
submitted changelists. This also adds support for the default changelist (for
all p4/p4d versions).
Reviewed at http://reviews.review-board.org/r/1031/
and http://reviews.review-board.org/r/804/
Fixes bug 1020.
|
|
|
|
|
|
|
|
This change introduces Repository UUID support to git-svn, allowing
Review Board to match repositories by UUID instead of comparing the
repository name directly.
Patch by Matthew Woehlke.
Reviewed at http://reviews.review-board.org/r/1030/
|
|
|
|
|
|
improperly check for the os.uname function.
Fixes bug #1280.
|
|
This introduces a new column, raw_file_url, to Repository, and shows it in the
admin UI for SCMTools that support it. This URL provides a way for certain
SCMTools (right now, Git only) to access files over HTTP, instead of checking
the file out through the SCM or accessing it locally (in the case of Git).
The raw file URL acts as a URL with fields to be substituted. If <revision>
exists in the URL, it will be replaced with the revision of the file. If
<filename> exists, it will be replaced with the filename.
This also adds a GitHub hosting service and bug tracker to the Repository
admin. Using this, it's trivial to add any GitHub project without knowing the
raw file URL mask.
Reviewed at http://reviews.review-board.org/r/977/
|
