[PATCH] Remove perl dependant commands if NO_PERL is set

[PATCH] Remove perl dependant commands if NO_PERL is set

[garga]

From: Renato Botelho <garga@FreeBSD.org>

git-submodule and git-request-pull are written in sh but use perl
internally. Add them to be replaced by unimplemented.sh when NO_PERL
flag is set
---
 Makefile | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/Makefile b/Makefile
index e326fa0..4dae0ec 100644
--- a/Makefile
+++ b/Makefile
@@ -1795,7 +1795,7 @@ git-instaweb: git-instaweb.sh GIT-SCRIPT-DEFINES
 	chmod +x $@+ && \
 	mv $@+ $@
 else # NO_PERL
-$(SCRIPT_PERL_GEN) git-instaweb: % : unimplemented.sh
+$(SCRIPT_PERL_GEN) git-instaweb git-submodule git-request-pull: % : unimplemented.sh
 	$(QUIET_GEN)$(RM) $@ $@+ && \
 	sed -e '1s|#!.*/sh|#!$(SHELL_PATH_SQ)|' \
 	    -e 's|@@REASON@@|NO_PERL=$(NO_PERL)|g' \
-- 
2.5.1

Re: [PATCH] Remove perl dependant commands if NO_PERL is set

[Junio C Hamano]

garga@FreeBSD.org writes:

> From: Renato Botelho <garga@FreeBSD.org>
>
> git-submodule and git-request-pull are written in sh but use perl
> internally. Add them to be replaced by unimplemented.sh when NO_PERL
> flag is set
> ---

Missing sign-off.

You also may want to hold off the "git-submodule" bit, as I expect
http://thread.gmane.org/gmane.comp.version-control.git/277128/focus=277129
would be one of the first changes to be in 'master' after the
upcoming release 2.6 (that is, it is likely that perl dependency
would be removed in 2.7).

>  Makefile | 2 +-
>  1 file changed, 1 insertion(+), 1 deletion(-)
>
> diff --git a/Makefile b/Makefile
> index e326fa0..4dae0ec 100644
> --- a/Makefile
> +++ b/Makefile
> @@ -1795,7 +1795,7 @@ git-instaweb: git-instaweb.sh GIT-SCRIPT-DEFINES
>  	chmod +x $@+ && \
>  	mv $@+ $@
>  else # NO_PERL
> -$(SCRIPT_PERL_GEN) git-instaweb: % : unimplemented.sh
> +$(SCRIPT_PERL_GEN) git-instaweb git-submodule git-request-pull: % : unimplemented.sh
>  	$(QUIET_GEN)$(RM) $@ $@+ && \
>  	sed -e '1s|#!.*/sh|#!$(SHELL_PATH_SQ)|' \
>  	    -e 's|@@REASON@@|NO_PERL=$(NO_PERL)|g' \

Re: [PATCH] Remove perl dependant commands if NO_PERL is set

[Renato Botelho]

> On Sep 3, 2015, at 16:11, Junio C Hamano <gitster@pobox.com> wrote:
> 
> garga@FreeBSD.org writes:
> 
>> From: Renato Botelho <garga@FreeBSD.org>
>> 
>> git-submodule and git-request-pull are written in sh but use perl
>> internally. Add them to be replaced by unimplemented.sh when NO_PERL
>> flag is set
>> ---
> 
> Missing sign-off.
> 
> You also may want to hold off the "git-submodule" bit, as I expect
> http://thread.gmane.org/gmane.comp.version-control.git/277128/focus=277129
> would be one of the first changes to be in 'master' after the
> upcoming release 2.6 (that is, it is likely that perl dependency
> would be removed in 2.7).

Can’t we add git-submodule to the list now, for 2.6, and then remove it when this change is pushed? This would help to have a sane tarball for 2.6. I’m considering this from a package maintainer perspective.

I’ll re-send it with sign-off.
--
Renato Botelho

[PATCH] Remove perl dependant commands if NO_PERL is set

[garga]

From: Renato Botelho <garga@FreeBSD.org>

git-submodule and git-request-pull are written in sh but use perl
internally. Add them to be replaced by unimplemented.sh when NO_PERL
flag is set

Signed-off-by: Renato Botelho <garga@FreeBSD.org>
---
 Makefile | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/Makefile b/Makefile
index e326fa0..4dae0ec 100644
--- a/Makefile
+++ b/Makefile
@@ -1795,7 +1795,7 @@ git-instaweb: git-instaweb.sh GIT-SCRIPT-DEFINES
 	chmod +x $@+ && \
 	mv $@+ $@
 else # NO_PERL
-$(SCRIPT_PERL_GEN) git-instaweb: % : unimplemented.sh
+$(SCRIPT_PERL_GEN) git-instaweb git-submodule git-request-pull: % : unimplemented.sh
 	$(QUIET_GEN)$(RM) $@ $@+ && \
 	sed -e '1s|#!.*/sh|#!$(SHELL_PATH_SQ)|' \
 	    -e 's|@@REASON@@|NO_PERL=$(NO_PERL)|g' \
-- 
2.5.1

Re: [PATCH] Remove perl dependant commands if NO_PERL is set

[Junio C Hamano]

Renato Botelho <garga@FreeBSD.org> writes:

>> You also may want to hold off the "git-submodule" bit, as I expect
>> http://thread.gmane.org/gmane.comp.version-control.git/277128/focus=277129
>> would be one of the first changes to be in 'master' after the
>> upcoming release 2.6 (that is, it is likely that perl dependency
>> would be removed in 2.7).
>
> Can’t we add git-submodule to the list now, for 2.6, and then remove
> it when this change is pushed? This would help to have a sane tarball
> for 2.6. I’m considering this from a package maintainer perspective.

Not really.  Anything brand-new that comes this late in the cycle
will not be in 2.6, so the earliest release this NO_PERL change can
appear in is one after the upcoming release.

Re: [PATCH] Remove perl dependant commands if NO_PERL is set

[Renato Botelho]

> On Sep 3, 2015, at 16:56, Junio C Hamano <gitster@pobox.com> wrote:
> 
> Renato Botelho <garga@FreeBSD.org> writes:
> 
>>> You also may want to hold off the "git-submodule" bit, as I expect
>>> http://thread.gmane.org/gmane.comp.version-control.git/277128/focus=277129
>>> would be one of the first changes to be in 'master' after the
>>> upcoming release 2.6 (that is, it is likely that perl dependency
>>> would be removed in 2.7).
>> 
>> Can’t we add git-submodule to the list now, for 2.6, and then remove
>> it when this change is pushed? This would help to have a sane tarball
>> for 2.6. I’m considering this from a package maintainer perspective.
> 
> Not really.  Anything brand-new that comes this late in the cycle
> will not be in 2.6, so the earliest release this NO_PERL change can
> appear in is one after the upcoming release.

Even if it’s a bug that produce a broken installation?

Nevermind, I’ll add the patch into FreeBSD ports so FreeBSD users can have a sane package installed.

--
Renato Botelho

Re: [PATCH] Remove perl dependant commands if NO_PERL is set

[Junio C Hamano]

garga@FreeBSD.org writes:

> From: Renato Botelho <garga@FreeBSD.org>
>
> git-submodule and git-request-pull are written in sh but use perl
> internally. Add them to be replaced by unimplemented.sh when NO_PERL
> flag is set
>
> Signed-off-by: Renato Botelho <garga@FreeBSD.org>
> ---
>  Makefile | 2 +-
>  1 file changed, 1 insertion(+), 1 deletion(-)
>
> diff --git a/Makefile b/Makefile
> index e326fa0..4dae0ec 100644
> --- a/Makefile
> +++ b/Makefile
> @@ -1795,7 +1795,7 @@ git-instaweb: git-instaweb.sh GIT-SCRIPT-DEFINES
>  	chmod +x $@+ && \
>  	mv $@+ $@
>  else # NO_PERL
> -$(SCRIPT_PERL_GEN) git-instaweb: % : unimplemented.sh
> +$(SCRIPT_PERL_GEN) git-instaweb git-submodule git-request-pull: % : unimplemented.sh
>  	$(QUIET_GEN)$(RM) $@ $@+ && \
>  	sed -e '1s|#!.*/sh|#!$(SHELL_PATH_SQ)|' \
>  	    -e 's|@@REASON@@|NO_PERL=$(NO_PERL)|g' \

Thanks, but this I suspect is insufficient.  In the pre-context of
your patch, you can see we have a rule to create git-instaweb when
NO_PERL is not in effect, so either way, we only have one rule to
create git-instaweb.

You are not disabling the rule to create the real git-submodule
and git-request-pull when NO_PERL is in effect with this patch,
without it, 'make' cannot tell which variant of git-submodule and
git-request-pull needs to be built.

I agree replacing commands with unimplemented may be a good thing; I
do not object to the goal of the patch.  But hopefully you now see
why it is too late to start discussing it with 2.6 as the goal.

Re: [PATCH] Remove perl dependant commands if NO_PERL is set

[Junio C Hamano]

Renato Botelho <garga@FreeBSD.org> writes:

>> Not really.  Anything brand-new that comes this late in the cycle
>> will not be in 2.6, so the earliest release this NO_PERL change can
>> appear in is one after the upcoming release.
>
> Even if it’s a bug that produce a broken installation?

Absolutely.

The way we do things around here is to do new features early in the
cycle before -rc0, fixes to pre-existing bugs before -rc1 and after
that we go to regression fixes only until the final.  And use of
Perl in submodule or request-pull is anything new in 2.6, I think.

This is a fix to pre-existing issue, and I haven't tagged -rc1 yet,
but speaking from experience, anything that I can poke a hole after
staring at it for only 20 seconds (see my other message about this
patch defining two different ways to build the same thing) has a
high chance of having other issues that will hurt other people in a
way other than the trivial "staring at it for 20-second" issue,
which would introduce a real regression.

Between an unknown regression and a known longstanding bug, we tend
to take the latter.

> Nevermind, I’ll add the patch into FreeBSD ports so FreeBSD users can
> have a sane package installed.

Sure.  That would help in the long term, as it will result in a
well-cooked fix in 2.7; by that time hopefully the 'submodule' bit
would not be necessary, but request-pull may still be using Perl.