GNU bug report logs - #59868
28.2.50; compilation-search-path incompatible with dir-locals

Previous Next

Full log

View this message in rfc822 format

From: Len Trigg <lenbok <at> gmail.com>
To: Eli Zaretskii <eliz <at> gnu.org>
Cc: 59868 <at> debbugs.gnu.org
Subject: bug#59868: 28.2.50; compilation-search-path incompatible with dir-locals
Date: Sat, 10 Dec 2022 08:11:10 +1300

[Message part 1 (text/plain, inline)]

On Fri, 9 Dec 2022 at 20:06, Eli Zaretskii <eliz <at> gnu.org> wrote:

> You are ignoring the first sentence in my response, right?
>

 I don't think so - the purpose of my email was essentially to ask why you
thought the variable shouldn't be buffer-local in *compilation* buffers.
I'm hoping to see where my mental model of using .dir-locals.el and/or
buffer-local variables is at odds with how emacs intends. The user manual
for both of these make it seem like this would be a perfect fit.


The rest was a suggestion (and I do see a good reason why project.el
> would like to provide this as a project-specific setting).
>
> More generally, why would you want a directory-specific value for this
> variable, if it isn't something the entire directory tree shares?
>

From my understanding, settings in .dir-locals.el *do* affect the whole
directory tree under the directory containing the file (and I'm happy for
that to be the case here). I think of .dir-locals.el in the project root as
the preferred way of specifying project-wide settings, and of
project.el/projectile.el as primarily providing meta functionality (e.g.
cross projects, or multi-file within a project). (I would also perhaps
naively expect that project.el would in turn get its project-wide settings
via that .dir-locals.el, at least that is how I have done it for projects
where I've needed to customize projectile settings (sorry, I haven't tried
project.el yet)).


Maybe you should tell why you need to tell Emacs where to find the
> sources?  Doesn't the tool you use to compile spell that out?
>

In my specific case the project is in R (which is a terrible language for
software development) and we are using a single top level makefile in the
project. It has a target to run linting, which (within each R linter
invocation) lints from a subdirectory, and the filenames in each lint
message are relative to those directories. There is not the hierarchy of
makefiles that you would often see in say C projects, where compilation
mode could parse make-issued messages about changing directories etc. It
seemed a perfect fit for compilation-search-path. I am currently working
around it with the following ugliness in my .dir-locals.el:

(
 (compilation-mode . ((eval . (progn
                                (setq-local compilation-search-path '(
                                              "~/sandboxes/myproject/R"
                                              "~/sandboxes/myproject/tests"
                                              ))
                                (put 'compilation-search-path
'permanent-local t)
                                (compilation-mode))))))


Cheers,
Len.

[Message part 2 (text/html, inline)]

Previous Next

GNU bug tracking system
Copyright (C) 1999 Darren O. Benham, 1997,2003 nCipher Corporation Ltd, 1994-97 Ian Jackson.