This is a partial revert of 1e20363.
When downloading a .diff.gz source package file, we do expect it to be
written to disk still compressed. If we were to uncompress it, then we
would get a size mismatch and even if we were to ignore that, we'd get a
hash mismatch.
On the other hand when downloading a changes file we need to make sure
that is written to disk uncompressed.
To make this work in both cases we can ask the HTTP server for no
special content encoding using "Accept-Encoding: identity". This is what
wget requests, for example. Then we can write the output to the file
without performing any decoding at our end by using the raw response
object again.
This fixes both cases.
LP: #2025748
```
ubuntutools/test/test_archive.py::LocalSourcePackageTestCase::test_pull
/usr/lib/python3.11/unittest/case.py:678: DeprecationWarning: It is deprecated to return a value that is not None from a test case (<bound method LocalSourcePackageTestCase.test_pull of <ubuntutools.test.test_archive.LocalSourcePackageTestCase testMethod=test_pull>>)
return self.run(*args, **kwds)
```
`test_pull` does not need to be run directly. Make it private.
Signed-off-by: Benjamin Drung <benjamin.drung@canonical.com>
pylint complains about C0209: Formatting a regular string which could be
a f-string (consider-using-f-string)
Signed-off-by: Benjamin Drung <benjamin.drung@canonical.com>
pylint complains about W1201: Use lazy % formatting in logging functions
(logging-not-lazy) and W1203: Use lazy % formatting in logging functions
(logging-fstring-interpolation).
Signed-off-by: Benjamin Drung <benjamin.drung@canonical.com>
Commit ae74f71a1e9d4be043162b19d23f2d44c964c771 removed the pylint unit
test saying that unit tests are not needed to just run flake8 or pylint.
Since pylint is useful, add it back, but this time call it directly and
not embed it into a unit test.
Signed-off-by: Benjamin Drung <benjamin.drung@canonical.com>
Use `--skip-patches` when extract source packages with `dpkg-source`.
`--no-preparation` is a source package build option and `--skip-patches`
is the correct extract option.
Signed-off-by: Benjamin Drung <benjamin.drung@canonical.com>
sponsor-patch calls `debdiff` which exits with 1 if there are
differences. So accept exit codes 0 and 1 as expected.
Signed-off-by: Benjamin Drung <benjamin.drung@canonical.com>
Fix operation of SourcePackage._source_urls() (as used, for example, in
SourcePackage.pull() called by backportpackage) to also work when the
class is instantiated with a URL as .dsc.
This is a regression caused by 1b12d8b4e3315de3bf417b40a3c66279f309d72c
(first in v0.184) that moved from os.path.join() to Pathlib, but
os.path.join() was also used to join URLs.
Thanks: Unit 193 for the initial patch.
Signed-off-by: Mattia Rizzolo <mattia@debian.org>
The last email announcing a 'hug day' appears to have been over
5 years ago, and the hugdaylist script doesn't seem to work
right anymore anyway.
This also removes ubuntutools/lp/libsupport, which has no functions
used by anything else.
Reading the raw stream doesn't decode files that are content-encoded,
which is true for the 'changes' file from launchpad, so it's saved
to disk gzipped which isn't what is expected.
Using the python requests iter_content method instead uses the
built-in stream decoding that the requests library provides, and
saves the file uncompressed/unencoded.
Note that since the encoded Content-Length won't match the resulting
unencoded actual length of data we save to file, this also turns off
the progress bar for any files that have Content-Encoding.
Yielding the result of super()_source_urls() / _binary_urls() yields the
generator object itself which generates the list of URLs - not the URLs
which would be returned from this generator. Instead use yield from which
forwards the yield onto the generator object itself.
Fixes LP: #1938659
Signed-off-by: Alex Murray <alex.murray@canonical.com>
Assumptions were made about the implementation by mocking
that are no longer true, and the tests generally need to be
fixed to be more robust about testing
Unfortunately private PPAs require downloading files from the special server
private-ppa.launchpad.net, and all the usual urls provided by the LP api fail.
So add code to handle using those custom private URLs, and use authentication
when downloading.
