Re: Dependencies should include "realpath"

[David Bremner <david@tethera.net>]

Ralph Seichter <ralph@ml.seichter.de> writes:

> * David Bremner:
>
>> Do you have a suggested replacement? I guess some inline perl with "use
>> Cwd 'realpath'" would probably work, although I haven't tested it.
>
> At a quick glance, that particular section of "configure" is run by
> doc/conf.py to generate three lines of Python code and store the result
> as sphinx.config, correct? If so, my preferred choice would be to use
> Python to figure out the absolute path, e.g. like so:
>
>   rsti_dir = os.path.abspath('emacs')
>
> This shows the generated result, and I assume that emacs is a directory
> in the source tree? I also wonder if an absolute directory path is really
> required for the doc-build to work.

Someone (TM) would have to check that this version did not break
out-of-tree builds. Other than that it seems plausible.

> The segment of conf.py which uses the generated config file does not
> look convincing to me anyway. Apparently the original author did not
> like it either, which is why the segment is labelled as "hacky". It
> should probably be overhauled, and not only because it uses the
> statement open(rsti_dir+'/'+file) which will potentially fail, depending
> on the build platform.

It's somewhat orthogonal to this discussion, but I don't object to a
patch replacing the hardcoded '/' with the appropriate python path
manipulation code. I think windows compatibility for notmuch has much
bigger issues, but it doesn't hurt to clean up the code.

Other than that, I'm not sure what an "overhaul" would involve.

d