Re: maint.mk sync

[Simon Josefsson]

Jim Meyering <address@hidden> writes:

> Simon Josefsson wrote:
>> I'm beginning to think that incremental syncing of maint.mk will never
>> finish, the sync targets are moving too quickly.
>>
>> I propose to just import the coreutils maint.mk into gnulib, and then
>> move forward from there on.  The patch below takes coreutils' maint.mk,
>> and keeps the current coverage+gettext+indent target, and use that in
>> gnulib.  I've verified that importing it into at least one of my
>> projects appears to work.  This could break some packages that use
>> maintainer-makefile.  I think the people who are using that have deep
>> understanding of these issues anyway, so they should be able to adapt.
>>
>> Ok to push?
>
> I've just looked through again and see several rules that people may
> find annoying: how many people care if you spell it "filesystem" or
> "file system"? ;-) It's easy to put up with that if the code already
> conforms, and even easier just to disable the check, but if a rule like
> that is turned _on_ by default, it'll just turn people off.

But the checks aren't turned on by default, are they?  The maintainer
needs to invoke the specific rules manually.

> Also, the diagnostic-checking rules are insufficiently general.  At the
> very least, they should take an optional cfg.mk-specified regexp to
> identify the functions used to emit diagnostics.

But it isn't used by default, either, is it?  So I don't think it is
harmful.

> One option is to maintain in maint.mk a list of the names of
> less-generally-useful rules, and disable those by default.
> Then we can leave most of the rules in the same file.

I'd like something like that: I think multiple projects will want these
less-generally-useful rules so it makes sense to share them.

The only thing that is missing is to make it easy for maintainers to put
something in cfg.mk that disables some checks they don't care about,
alternatively list the checks they do care about.

Hm.  Possibly the 'syntax-check-rules' variable should only be set
through '?=' Then every maintainer can filter out some checks they don't
care about.

>> I know there are some coreutils specific stuff below, but they don't
>> seem to cause harm.  The intention is that this can be fixed later.
>>
>> When this has been pushed into gnulib, I'll send a patch to coreutils
>> that makes it use maintainer-makefile from gnulib.  It should be quite
>> short.
>
> You're welcome to push that, as far as I'm concerned.

I'll wait a few hours more, but I'll test the coreutils maint.mk with my
projects meanwhile.  I don't think it will cause any problem.

/Simon