On Tue, Jan 1, 2013 at 11:21 PM, Jonathan Nieder <jrnieder@xxxxxxxxx> wrote:
>
> How about this patch (untested)?
Looks good. Thanks!
>> --- a/Makefile
>> +++ b/Makefile
>> @@ -2267,12 +2267,9 @@ $(patsubst %.py,%,$(SCRIPT_PYTHON)): % : unimplemented.sh
>> mv $@+ $@
>> endif # NO_PYTHON
>>
>> -configure: configure.ac GIT-VERSION-FILE
>> +configure: configure.ac
> [...]
>> --- a/configure.ac
>> +++ b/configure.ac
>> @@ -142,7 +142,10 @@ fi
>> ## Configure body starts here.
>>
>> AC_PREREQ(2.59)
>> -AC_INIT([git], [@@GIT_VERSION@@], [git@xxxxxxxxxxxxxxx])
>> +AC_INIT([git],
>> + m4_esyscmd([ ./GIT-VERSION-GEN &&
>> + { sed -ne 's/GIT_VERSION = //p' GIT-VERSION-FILE | xargs echo -n; } ]),
>> + [git@xxxxxxxxxxxxxxx])
>
> I don't think that would warrant dropping the GIT-VERSION-FILE
> dependency, since the resulting configure script still hard-codes the
> version number.
Yeah, you're right. I was merely sweeping the dependency under the rug :-(
>
> diff --git a/Makefile b/Makefile
> index 736ecd45..2a22041f 100644
> --- a/Makefile
> +++ b/Makefile
> @@ -2275,7 +2275,7 @@ configure: configure.ac GIT-VERSION-FILE
> $(RM) $<+
>
> ifdef AUTOCONFIGURED
> -config.status: configure
> +config.status: configure.ac
> $(QUIET_GEN)if test -f config.status; then \
> ./config.status --recheck; \
> else \
The next line just outside the context here does depend on
'configure', which is why I thought this would not be right. But it
seems impossible to get away from that, and AUTOCONFIGURED should only
be set when ./configure has been run (IIUC), so it's not even
realistic to have "git reconfigure" fail to find "./configure". So,
again, looks good.
--
To unsubscribe from this list: send the line "unsubscribe git" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at http://vger.kernel.org/majordomo-info.html