Re: pcase is autoloaded?

[Aidan Gauland <aidalgol@no8wireless.co.nz>]

Stefan Monnier <monnier@iro.umontreal.ca> writes:

>> This means that people leave it to be autoloaded rather than
>> specifically required. In our case it broke our test badly because our
>> file-system mock was loading pcase.el (which didn't work, surprise!)
>
> I have no idea what problem you bumped into.  I don't know what is your
> "file system mock", nor do I know why loading pcase.el didn't work (or
> why it should come as (no?) surprise).
>
>> Is this a good idea? I don't think so.
>
> Please give more concrete details,

Here is the relevant backtrace: https://gist.github.com/aidalgol/5070012

If pcase is not loaded when the test is run, this happens; if pcase *is*
loaded when the test is run, this error does not occur.  This does seem
to suggest that the autoloading of pcase causes a problem, but as Sacha
Chua pointed out (on #emacs), `fakir--expand-file-name' does not
correctly handle the a nil case, and calls `file-name-as-directory' with
nil.  (I do not know how that case should be handled.)

From what I can tell, pcase being autoloaded is not the problem, it's
just exposing a bug in fakir, and explicitly loading pcase before
running the above test only masks the problem.

Regards,
Aidan Gauland