| Age | Commit message (Collapse) | Author |
|---|
|
There have been rather obvious errors in there, as we forgot to unregister
the filehandles. Now we will read from a buffer ourselves, which should be
faster and ideally, doesn't lead to spurious errors anymore.
Related to #232
|
|
Additionally, for this and future examples, there is a test_doc.py
suite to contain all code mentioned in the docs. That way, we know
if things stop working.
Fixes #236
|
|
Fixes #235
[ci skip]
|
|
Assured unicode values are supported when reading the configuration,
and when getting author/committer information from the environment.
Fixes #237
|
|
Once again, __del__ is not being called as usual anymore in python 3 ... .
|
|
path
Fixes #238
|
|
A test verifies it's truly working.
Related to #238
|
|
Improved a test to assure multiple update(init=False|True) calls don't
throw.
Fixes #93
|
|
After all, it was easier than expected. It seems that previous assertions
the test made should have never been true to begin with. Thus we might
have improved the test thanks to our improved implementation.
Fixes #233
|
|
This also means that now we seem to be able to properly handle .git files in submodules
Related to #233
|
|
The root-submodule test is still failing though, this time even earlier
than before
|
|
A simple test verifies this at least.
|
|
However, a simple test-case still fails for reasons not yet understood.
There is more to be fixed here - .remove() still fails.
|
|
There is some more work to do, as renames and updates still have to be
adjusted accordinlgy.
Relates #233
|
|
It actually revealed a bug in the implementation of Submodule.add,
which just showed in python 3 for the wrong reasons. Thankfully,
failing tests after all allowed to get this issue fixed ... .
|
|
git-python.
This is the case with `git submodule add` as well. This makes sense as
an empty git repository doesn't have a commit, which needs to be specified
as SHA in the parent repositories tree entry for the respective submodule.
When manually adding the empty submodule to the .gitmodules file, git-python
will throw another error related to the inability to find the submodule
in the index.
Even if an iteration would be possible, git-python would now throw
a BadName exception, which clearly indicates that the 'HEAD' revision
is invalid (as it doesn't point to any commit).
Fixes #152
Fixes #105
|
|
string.
Previously, it would say it can handle absolute module paths, but didn't
actually do so.
A test-case was improved to check for this case.
Fixes #161
|
|
Adjusted code to not check for .gitmodules existence anymore, we will
deal with it.
Fixes #117
|
|
This is verified by the respective test.
Fixes #117
|
|
Fixes #7
[ci skip]
|
|
It does not necessarily seem to work, but at least we don't access a
dict concurrently anymore.
|
|
This allows sub-trees to be archived as well, and makes `.archive()`
feature complete.
Fixes #67
|
|
We implement it as described in this article:
http://stackoverflow.com/questions/1557183/is-it-possible-to-include-a-file-in-your-gitconfig
Thus we handle
* cycles
* relative and absolute include paths
* write-backs in case of writable GitConfigParser instances
Fixes #201
|
|
Fixes #169
|
|
However, it does so only on posix. The test-case will run on posix only
as well.
Please note that in theory, even on windows we will attempt to run hooks,
even though I am not sure that this will actually work.
Fixes #81
|
|
used consistently.
This will save IOPs, and make the code easier to understand (I suppose).
Related to #224
|
|
required afterall.
Turns out that the now removed `read_gitfile` functions applied os.path.realpath which caused
the issue described in #224.
Fixes #224
|
|
index related work.
That way, we don't try to compare a real-path to a non-real one, which would make the implementation think
a file is not actually part of the repository.
Fixes #224
|
|
For some reason, I didn't trust the existing one as it tests that code
more indirectly.
Related to #160
|
|
The point is that XDG_CONFIG_HOME is meant to point at the '.config' directory,
whereas '.config' has to be added only if HOME is used instead.
Fixes #160
|
|
Fixes #160
|
|
Related to #83
|
|
general.
Fixes #83
|
|
Now by default, we will not walk up the directory structure and possibly find
git directories that the user didn't intend to find.
If required, that kind of behaviour can be turned back on.
Fixes #65
|
|
That way, flake won't have to completely ignore issues it might have
in other portions of this file.
|
|
test-file is working.
This brings us much closer to what git can do, and should at least prevent errors while reading
configuration files (which would break a lot of features, like handling of remotes since these rely
reading configuration files).
Fixes #112
|
|
It works as expected by me at least.
Related to #71
|
|
Additionally, unicode handling was improved to the point where we deal
with all diff(create_path=True) data as binary.
Therefore we don't claim to know all encodings of all textfiles in the world,
even though we still assume that everything git throws at us is utf-8 encoded.
Fixes #113
|
|
future
There is still some work todo in terms of how we handle the encoding
|
|
Related to #74
|
|
Related to #74
|
|
Fixes #74
|
|
Now handling the new exception BadName as well
|
|
|
|
Helps to clarify #55
|
|
Previously, an unresolvable ref name like HEAD would end up as HEX and
was presented as BadObject error, even though that exception is for
invalid shas only.
Now BadName is thrown, which converts into a more useful error message.
Improves #105
|
|
Forgot to fix the test, which used the same broken version_info condition
|
|
Apparently, thanks to an incorrect version check, PY3 ended up using
a git command object database by default. This is now fixed.
Additionally, the update_cache code was adjusted to check for method-existence,
as it's valid to use object databases which simply don't have a
caching mechanism (like the git command object database)
|
|
Fixes #34
|
|
See #47
|
