[Toybox] [PATCH] Try to debug modinfo test flake.

[Rob Landley]

On 6/7/23 14:40, enh wrote:
> indeed. i don't know _why_ this is happening,

Race condition setting up the VM? Permissions hiccup? No idea, but I've been
rereading the Discworld "watch" books recently and... let's just say by halfway
through Jingo prolonged exposure to Sam Vimes is not making me LESS inclined to
track things down. :P

In the "I break everything" department: your description has a patch in it, so
"git am" on the attached file is barfing because the file "actual" is not in the
repo. (I would have assumed git format-patch had an existing escape syntax for
the in-band signalling the way "From:" lines in mbox files do, but apparently not?)

> but i think that means the fix (attached) is obvious at least.

Your sort returns modules from the newest kernel installed, but I let apt-get
upgrade install new kernels regularly-ish, but sometimes don't reboot my laptop
for months at a time, so the newest kernel installed in the filesystem and the
one I'm running aren't necessarily the same thing. (And sort isn't version
sort...) Also, in _theory_ you could have one module each installed from two
kernel versions (but a _different_ single module installed from each kernel) and
then the two modules test would fail...

Instead I'm adding $(uname -r) to $MODULE_ROOT in the non-android case to avoid
duplicates, and... why are we discarding stderr on the calls to head and tail?
(It's not going to discard it from find, it's in the wrong place in the pipeline
for that...)

I also added comments about why we DON'T have to quote stuff, because it was
either that or add the quotes to get my shoulders to relax...

Try commit a5b5749cc7f0

Rob