James Youngman wrote:
> On Mon, Nov 9, 2009 at 12:46 PM, Eric Blake <
ebb9@...> wrote:
>> -----BEGIN PGP SIGNED MESSAGE-----
>> Hash: SHA1
>>
>> According to James Youngman on 11/9/2009 3:09 AM:
>>> On Mon, Nov 9, 2009 at 9:44 AM, Jim Meyering <
jim@...> wrote:
>>>
>>>> [btw, the above check for unmarked-yet-translatable diagnostics is automated
>>>> in gnulib's maint.mk. If you're interested, running "make syntax-check"
>>>> can detect problems like this; to enable, use gnulib's maintainer-makefile
>>>> module. ]
>>>
>>> Maybe it's useful to combine this with switching to tracking gnulib as
>>> a submodule instead of via import-gnulib.conf. Eric, WDYT?
>>
>> Very much in favor - git submodule is more mature now, and is definitely a
>> nicer way of tracking exactly which version of gnulib we depend on.
>
> Good, let's go ahead.
>
> I'm sure there will be code that "make syntax-check" doesn't like (for
> example, whitespace at the ends of lines I'd assume). So I'm sure
> we'll need to adopt some kind of phased approach. I'm happy to do a
> significant share of that kind of work. However, if you're willing
> I'd be happy for you to do the submodule change, since you've clearly
> done it before.
Good to hear it!
As an example of how to do that, see gzip.git.
I added this to its cfg.mk (included by GNUmakefile), since
for now the listed tests would fail, and fixing them wasn't a priority:
# Tests not to run as part of "make distcheck".
local-checks-to-skip = \
sc_error_message_period \
sc_error_message_uppercase \
sc_m4_quote_check \
sc_obsolete_symbols \
sc_program_name \
sc_prohibit_S_IS_definition \
sc_prohibit_atoi_atof \
sc_prohibit_stat_st_blocks \
sc_space_tab \
sc_useless_cpp_parens
To stage things, just add the name of each initially-failing test
to that list, and as you find the time/desire to fix things, remove
each rule name when the code is ready to pass the corresponding test.
