| Commit message (Collapse) | Author | Age |
|
|
|
|
|
|
|
|
| |
Git itself doesn't care but Python wants a bytestring. This only matters
when using the C locale.
We delay conversion so we don't have to handle fallback_author
separately. We can't convert unconditionally since pq-rpm is allowed to
pass in a None author.
|
|
|
|
| |
Closes: #761166
|
|
|
|
|
|
| |
if the pq branch doesn't exist yet.
Closes: #876800
|
|
|
|
|
|
|
| |
This would strip away text in []. This improves the import/export round
trip.
Closes: #872354
|
|
|
|
| |
The output can be in any encoding so assuming utf-8 is plain wrong
|
| |
|
|
|
|
| |
broken by 75e18fc5bb4831df924bc708130b0a6ab311a230
|
|
|
|
| |
Use present instead of past tense
|
|
|
|
|
|
| |
We always want to start from an imported series.
Thanks: Chris Lamb for the reproducer
|
| |
|
|
|
|
|
| |
We skipped the commit when the patch series became completely empty
but we want to commit that change too.
|
|
|
|
|
|
|
|
| |
So far we would fail empty commits with the confusing error
$ gbp pq export --commit
gbp:info: Generating patches from git (debian/sid..patch-queue/debian/sid)
gbp:error: Error running git commit:
|
| |
|
| |
|
| |
|
| |
|
| |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
to be specified as an option
Now that gbp-pq preserves patch names on import and export the
--[no-]patch-numbers options have been rendered largely useless unless
the patch queue commits are amended with the Gbp[-Pq]: Name tags
removed. The options remain useful if support for renumbering is
added. It's also handy to allow the choice of the format to use
for the numeric prefixes.
Closes: #790890
Signed-off-by: Guido Günther <agx@sigxcpu.org>
|
|
|
|
| |
use gnu.org/licences instead of a postal address, as suggested by Guido
|
|
|
|
| |
Closes: #761161
|
| |
|
|
|
|
| |
Remove unused imports, remove unused variables or actually use them
|
|
|
|
| |
The former is supported with python3
|
| |
|
|
|
|
|
|
|
|
|
| |
When we write out patches to subdirs using a topic we want to filter out
this topic from the commit message. Support for this was lost in
7ce15d2434ee42aa5a1afce3d03069c5efb2db1b
add it back. Also fix parsing of the deprecated commands.
|
|
|
|
| |
We shouldn't generate the deprecated ones
|
|
|
|
|
|
| |
So that it can be re-used by the upcoming pq-rpm tool.
Signed-off-by: Markus Lehtonen <markus.lehtonen@linux.intel.com>
|
|
|
|
| |
Closes: #761160
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
Use our own function for constructing the patch files instead of using
the format-patch command of git. This way, we get the desired output
format directly, without the need for the error-prone "format-patch,
parse patch files, mangle and re-write patch files" cycle.
Also, fix patch naming in patch generation when '--no-patch-numbers' is
used. Previously, multiple commits with the same subject resulted in
multiple patches having the same filename. This lead into broken series
with missing patches as patch files were overwritten by the topmost
commit.
Signed-off-by: Markus Lehtonen <markus.lehtonen@linux.intel.com>
|
|
|
|
| |
and better use the output for verification
|
|
|
|
|
|
|
| |
This puts test dirs below /tmp which often is a tmpfs.
All tests include the context module which consolidates tmpdir creation
and cleanup, undoes a chdir in teardown and silences log messages.
|
|
|
|
|
|
|
|
|
| |
Call the author parsing/guessing function outside the apply patch
functions. This way, the caller can decide when to do the guessing, and
with which parameters. Now the apply_patch functions do what their name
suggests.
Signed-off-by: Markus Lehtonen <markus.lehtonen@linux.intel.com>
|
| |
|
| |
|
|
|
|
|
| |
so the RPM based tools don't need to rely on a control file but
can e.g. look at the spec file.
|
|
|