snap: add support for parsing snap layout section

[zyga]

The layout section defines the "shape" of the runtime mount namespace
of a given snap. Layout is entirely optional and is empty by default.

Tree kind of layout entries can be defind:

• bind mounts, designated by the presence of the "bind" field
• tmpfs mounts, designated by the "type" field of "tmpfs"
• symlinks, designated by the presence of the "symlink" field

In addition, all kinds can use the "user", "group" and "mode" fields
to define the ownership and permissions of any parent directories that
may need to be created. At present those are limited to "root", "nobody"
and at most "0755" permissions.

Any path designators can expand the following environment variables:
$SNAP, $SNAP_DATA and $SNAP_COMMON.

Forum: https://forum.snapcraft.io/t/feature-snap-layouts/1471
Signed-off-by: Zygmunt Krynicki zygmunt.krynicki@canonical.com

[codecov-io]

Codecov Report

Merging #3636 into master will increase coverage by 0.03%.
The diff coverage is 91.56%.

@@            Coverage Diff             @@
##           master    #3636      +/-   ##
==========================================
+ Coverage   75.32%   75.36%   +0.03%     
==========================================
  Files         389      389              
  Lines       33666    33748      +82     
==========================================
+ Hits        25359    25434      +75     
- Misses       6489     6494       +5     
- Partials     1818     1820       +2

Impacted Files	Coverage Δ
snap/info.go	88.62% <ø> (ø)	⬆️
snap/info_snap_yaml.go	94.05% <87.5%> (-0.5%)	⬇️
snap/validate.go	95.39% <93.22%> (-1.38%)	⬇️
interfaces/sorting.go	98.71% <0%> (-1.29%)	⬇️
cmd/snap/cmd_aliases.go	95% <0%> (+1.66%)	⬆️

---

Continue to review full report at Codecov.

Legend - Click here to learn more
Δ = absolute <relative> (impact), ø = not affected, ? = missing data
Powered by Codecov. Last update 1398214...0faf6fd. Read the comment docs.

[mvo5]

Code looks good from my POV, I had a hunch that VlidatePathVariables() could be written in a simpler way but I played around a bit and my version was not easier to read so my hunch was probably wrong.

[chipaca]

mostly ok, but needs a fix

[chipaca]

mode_t is unsigned, and os.FileMode is a uint32; math.MaxUint32 is a valid os.FileMode, and would fail to parse here. But also, os.ModeDir, a constant pre-defined probably-not-that-rare file mode, is not a valid int32.

[zyga]

Done, I think.

[chipaca]

nope, you're still using ParseInt instead of ParseUint

[zyga]

Ah, I didn't notice that the parsing function is incorrect too. Thank you!

[chipaca]

interesting that we don't support having a $ that isn't interpolating a variable (i.e., no \$)

[zyga]

Yeah, I kind of think this is a feature though :)

[chipaca]

// ValidateLayout does unspeakable things to the layout

[zyga]

ValidateLayout makes someone's hair gray ;-)

[zyga]

Documented

[chipaca]

thank you