Integrate component based path layering

[cmacq2]

Integrate component based path layering into the main brickstrap script itself.

With this change brickstrap can take advantage of component base path scheme discussed in issue #17 and implemented in PR #22. For details refer to: Discussion: https://github.com/ev3dev/brickstrap/issues/17 Implementation PR: https://github.com/ev3dev/brickstrap/pull/22

Additional changes:

• Namespace functions in brickstrap.sh
• Move 'inline' code into proper standalone functions where possible
• Fix -h switch
• Enable other login shells besides bash

This change does not integrate the architecture/QEMU support from PR #21

[cmacq2]

Updated to integrate feedback. I've tested this with a fresh run of a project build (using 'all') to check that the function renaming didn't accidentally break things.

[cmacq2]

Do you have any further comments? Or can the PR be merged in as it stands now?

[dlech]

I would like to actually test it myself first. Haven't got around to that yet.

[cmacq2]

Okay, looking forward to test results! Tip: pass -c . if you haven't converted your project directories yet.

[cmacq2]

I've tightened validation of project & component names. This change disallows the following 'project' paths:

• brickstrap source directory
• tests/ subdirectory of brickstrap
• docs/ subdirectory of brickstrap

Additionally trivially duplicate component names are now ignored, and component names which attempt to 'escape' the project directory like ../ are disallowed.

There's still the edge case of someone specifying a non-trivial duplicate component name, for example: -c name1/../name2 -c name2

[cmacq2]

Bug fix: match scripts against whole lines from preinst blacklist; substring matches should no longer lead to false positives.

Second attempt: turns out grep supports the anchors anyway, and using plain grep avoids escaping issues w.r.t. regex syntax.

[cmacq2]

Updated to use the fgrep -xq "<pattern>" construct. This is cleaner than the previous 'fixes'..

[dlech]

Shouldn't we be using fgrep everywhere else we are grepping a variable string?

[cmacq2]

Unless we explicitly use the anchors... fgrep/grep -F doesn't do those. I'll prepare a new update. I checked: it's just brickstrap.sh that needs to be adjusted.

[dlech]

Thanks for fixing up the greps. I'm running it one more time to make sure that changing the greps didn't break anything. In the meantime, the readme needs to be updated for the command line switch changes.

[cmacq2]

I'd like to postpone the readme until we've also integrated the architecture/QEMU related changes into brickstrap.sh.

Also: if you are going to develop issue #24 further that will require additional matching readme updates as well.

[dlech]

Thanks!