Hard-link local sources instead of symlinking them.

[kyrofa]

This PR fixes LP: #1590108 by hard-linking local sources instead of symlinking them. If hard links aren't possible, it copies them instead.

[kyrofa]

Take particular note here: previously the code was very careful not to overwrite an existing parts/<name>/src directory or file. That logic is essentially impossible now that we're not symlinking, so I decided to assert ownership over parts/<name>/ and just blow it away if it already exists. See any downside to that? Users shouldn't be putting things there...

[sergiusens]

ownership was always the case.
We do need to track source changes now though ;-)

[kyrofa]

We do need to track source changes now though ;-)

I'm not sure what you mean. Are you saying something needs to happen here?

[sergiusens]

El 01/07/16 a las 18:25, Kyle Fazzari escribió:

In snapcraft/internal/sources.py
#622 (comment):

         os.remove(self.source_dir)

•    elif (os.path.isdir(self.source_dir) and
  

•          not os.listdir(self.source_dir)):
  

•        os.rmdir(self.source_dir)
  

•    elif os.path.exists(self.source_dir):
  

•        raise EnvironmentError('Cannot pull to target {!r}'.format(
  

•            self.source_dir))
  

•    elif os.path.isdir(self.source_dir):
  

•        shutil.rmtree(self.source_dir)
  

We do need to track source changes now though ;-)

I'm not sure what you mean. Are you saying something needs to happen here?

Eventually, when we implement tracking of changes to sources that would
affect build results ;)

[kyrofa]

Ah, yes of course. That will be a wonderful day! This change brings it a step closer 😃 .

[sergiusens]

please recreate an integration test with this scenario to confirm this

[kyrofa]

Done.

[sergiusens]

ah, we might be missing .snapcraft.yaml here!

[kyrofa]

Ah, indeed we are! I'll go ahead and fix that here.

[kyrofa]

Done.

[sergiusens]

looks good, I am reaching EOD so will let @elopio pay attention to the detail in UTs ;-)

[sergiusens]

parts/<part-name>/bin/pkg-config should be erased from existence 😉

[sergiusens]

just that last comment from me :-)

[kyrofa]

parts/<part-name>/bin/pkg-config should be erased from existence

Done 😉 .

[kyrofa]

Huh... now that I merged in master suddenly the coverage is decreasing instead of increasing. Not sure what the deal is there. Also, even VPNd in I can't seem to get to the autopkgtest failure, though judging from previous results it's just ROS.

[kyrofa]

Huh. Now coverage increased again. Coveralls I think you're full of it.

[kyrofa]

Failing ROS:

ValueError: Invalid placeholder in string: line 1, col 123
\x1b[31m\x1b[1m<==\x1b[0m Failed to process package \'\x1b[1m\x1b[34mtalker\x1b[0m\': 
  Invalid placeholder in string: line 1, col 123\x1b[0m
Command failed, exiting.
\x1b[0;31mCommand \'[\'/bin/sh\', \'/tmp/tmpm099_p1z\', \'/bin/bash\', \'/tmp/tmpzbdd1lf3\']\' returned non-zero exit status 1\x1b[0m
'}}}

Not really sure what the deal is with autopkgtests...

[elopio]

This is all green. The examples error is in the cleanup, fixed here: ubuntu-core/snappy-jenkins#193