Methods for Improving Build Tools

[Michael Stone]

Joel,

> What kind of improvement did you have in mind? 

Two things: 

 * First, get software and instructions that allow one to get
   N images mounted simultaneously so that images can be more easily
   diffed. Or that allows one to change which image is mounted after
   copying it to a filesystem.
 
 * Second, write a script that adds writable blocks (filled with
   0xFF) to an image so that modifications can be made to the mtd and
   propagated to the underlying block device (i.e. the loopback device).

> Is the task important enough to have a solution such that an average
> Fedora/Ubuntu system could mount a JFFS2 image?

Not sure what this question is asking. If it's asking "how important is
this?", then my answer is "the world won't end if it doesn't happen, but
it would certainly be convenient if we had it".

> To get there, is it a matter of refining the instructions and perhaps
> creating a bundle of the relevant scripts and software? Or does it
> require the development of something new?

It's mostly instructions and script-writing. Basically, you could think
of it as "improvements to the mtd-utils package".

> I don't (yet) have a deep understanding of filesystems nor how the
> FUSE works, but would writing a FUSE driver that could mount a jffs2
> image be a suitable solution? I am assuming this is possible.

It could be done, but advantage would there be over using the
loopback+mtd2block technique that dwmw2 already suggested?

> I would like to take up this task. Would it be of a suitable scope for
> a Summer of Code project?

My feeling is that for someone like yourself, it's more like a three
week project rather than a ten week project, but there might be a few
other things it could be combined with that would make it more
appropriate in scope.

Cheers,

Michael