Use a recursive iglob for filesets

[sergiusens]

This adds support for ** in filesets for parts.

LP: #1616464

Signed-off-by: Sergio Schvezov sergio.schvezov@ubuntu.com

[come-maiz]

retest this please

[come-maiz]

Please copy the comment:

Remove .la files which don't work when they are moved around

[come-maiz]

👍
I worry about @kyrofa, he was so happy about the other branch.

[sergiusens]

El 30/08/16 a las 00:49, Leo Arias escribió:

👍
I worry about @kyrofa https://github.com/kyrofa, he was so happy
about the other branch.

He said he was going to be happy with this one as well ;-)

[kyrofa]

If this fails, it might be much easier to digest an error based on the length of the lists rather than directly comparing their contents (it'll print the lists, right?). You could still print the list in the failure message if you wanted.

[kyrofa]

Very nice! I made one ignorable suggestion, 👍 from me.

[ghost]

So this is actually a problem. Specifically with the .pth removal. There are several projects that I know of that have the .pth to (ab)use python importing. Here is a line from the dogpile.cache pth file that allows it to import from site-packages/dogpile/cache without having site-packages/dogpile/init.py

import sys, types, os;p = os.path.join(sys._getframe(1).f_locals['sitedir'], *('dogpile',));ie = os.path.exists(os.path.join(p,'__init__.py'));m = not ie and sys.modules.setdefault('dogpile', types.ModuleType('dogpile'));mp = (m or []) and m.__dict__.setdefault('__path__',[]);(p not in mp) and mp.append(p)

there are many such projects that do things like this unfortunately.

[sergiusens]

El 22/09/16 a las 13:44, Sam Yaple escribió:

So this is actually a problem. Specifically with the .pth removal.
There are several projects that I know of that have the .pth to
(ab)use python importing. Here is a line from the dogpile.cache pth
file that allows it to import from site-packages/dogpile/cache without
having site-pacakges/dogpile/init.py

import sys, types, os;p =
os.path.join(sys./getframe(1).f_locals['sitedir'], _('dogpile',));ie =
os.path.exists(os.path.join(p,'_init/.py'));m = not ie and
sys.modules.setdefault('dogpile', types.ModuleType('dogpile'));mp = (m
or []) and m.dict.setdefault('_path*',[]);(p not in mp) and mp.append(p)

there are many such projects that do things like this unfortunately.

In your snapcraft.yaml, can you try to explicitly add it?

[ghost]

Thats a silly thing to suggest. This was rather complicated to even figure out why it was broken. Even though I can add this file specifically in, this will lead many people to believe that python does not work well with snapcraft.

Indeed, looking through the IRC logs show several cases where people are complaining about it not being able to import a package after this patch merged, I would bet at least some of those are this exact issue.

Why is removing the pth file something that is even needed?

[sergiusens]

El 22/09/16 a las 14:25, Sam Yaple escribió:

Thats a silly thing to suggest. This was rather complicated to even
figure out why it was broken. Even though I can add this file
specifically in, this will lead many people to believe that python
does not work well with snapcraft.

I am not saying we should keep it this way! :-)

Indeed, looking through the IRC logs show several cases where people
are complaining about it not being able to import a package after this
patch merged, I would bet at least some of those are this exact issue.

Why is removing the pth file something that is even needed?

I think this was just oversight

[ghost]

Submitting a patch :)

Not a big deal in the grand scheme of things. Shame i just missed 2.18 window though

[sergiusens]

El 22/09/16 a las 14:29, Sam Yaple escribió:

Submitting a patch :)

Not a big deal in the grand scheme of things. Shame i just missed 2.18
window though

don't worry, I am preparing 2.18.1 right now.