This can occur e.g. if directory permissions prevent one from writing to
the local cloud storage directory. (Such a crash was discussed on
mailing list.) The error message on GUI is misleading, claiming that
cloud connection failed...
Signed-off-by: Miika Turkia <miika.turkia@gmail.com>
---
We probably should check the return value of other git operations as
well, but going for bare minimum for now.
The following pragma is Clang specific:
It produces a warning:
warning: ignoring #pragma clang diagnostic [-Wunknown-pragmas]
Only enable it for Clang by checking the __clang__ macro.
Signed-off-by: Lubomir I. Ivanov <neolit123@gmail.com>
Signed-off-by: Dirk Hohndel <dirk@hohndel.org>
On Windows that would fail because stat() doesn't deal well with our
utf8 strings.
Added new subsurface_stat() portability function to replace stat().
Added Windows implementation of subsurface_stat() using wstat(),
with conversion to ut16 of the inputed path.
Other platform implementations (linux, android) make use of the normal stat().
Added non ASCII test case in TestGitStorage::testGitStorageLocal()
Signed-off-by: Jeremie Guichard <djebrest@gmail.com>
The authentication count was a static counter in the authentication
callback, which gets incremented until we consider the authentication to
have failed.
The problem with that is that it doesn't get incremented for just _one_
authentication operation, it gets incremented each time you try to load
or save, so eventually the code considers authentication to have failed
even if nothing ever went wrong.
This fixes it by making it static to the whole git-access file, and have
each operation clear it before starting a new remote access.
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Signed-off-by: Dirk Hohndel <dirk@hohndel.org>
So far we didn't do that at all, we either relied on the user manually
creating a local repo, or we cloned a remote repo.
Signed-off-by: Dirk Hohndel <dirk@hohndel.org>
If we have incorrect cloud credentials, we need to return an error on
git authentication call back in order to avoid endless authentication
loop. This might well happen e.g. when changing the password on desktop
and then on laptop Subsurface still thinks the credentials are validated
and ends up in the authentication loop.
The authentication call back on libgit is intended to be used to ask for
user credentials, and as we handle credentials elsewhere, we just need
to fail the authentication attempts. (The threshold for bail out could
have been 1 attempt...)
Signed-off-by: Miika Turkia <miika.turkia@gmail.com>
Signed-off-by: Dirk Hohndel <dirk@hohndel.org>
Having subsurface-core as a directory name really messes with
autocomplete and is obviously redundant. Simmilarly, qt-mobile caused an
autocomplete conflict and also was inconsistent with the desktop-widget
name for the directory containing the "other" UI.
And while cleaning up the resulting change in the path name for include
files, I decided to clean up those even more to make them consistent
overall.
This could have been handled in more commits, but since this requires a
make clean before the build, it seemed more sensible to do it all in one.
Signed-off-by: Dirk Hohndel <dirk@hohndel.org>
