You are viewing a plain text version of this content. The canonical link for it is here.
Posted to dev@apr.apache.org by Rainer Jung <ra...@kippdata.de> on 2010/07/26 12:30:17 UTC

Backporting and forward porting on APR / APR-UTIL

I'm a bit undecided whether to port some changes between APR and 
APR-UTIL branches:

- r780882 (wrowe): fix vpath building for xml/expat (removing 
"configure" target in Makefile.in)
   The change is in the 1.3.x branch, but neither in any older nor newer 
branches.

- r979102 (rjung): Updating buildconf.sh and configure.in for in 
xml/expat to allow building with libtool 2. Applied to 1.5, 1.4 and 1.3. 
Should I backport to older branches, or do we want to stay very 
conservatve here? I'd say backport, because otherwise we might run into 
the same problem we had with the first tarball for 2.2.16 again, namely 
that the machine of the RM is to modern, even when preparing e.g. an 
httpd 2.0 release.

- r979109 (rjung): Updating config.guess and config.sub to allow for 
better detection of modern platforms. Applied to APR 1.3, ..., trunk and 
apr-util 1.3, ..., 1.5. I suggest applying to the old branches as well.

Regards,

Rainer


Re: Backporting and forward porting on APR / APR-UTIL

Posted by Rainer Jung <ra...@kippdata.de>.
All ports done (together with your r979391).

Re: Backporting and forward porting on APR / APR-UTIL

Posted by Rainer Jung <ra...@kippdata.de>.
On 26.07.2010 17:32, William A. Rowe Jr. wrote:
> On 7/26/2010 10:10 AM, Rainer Jung wrote:
>> On 26.07.2010 15:47, William A. Rowe Jr. wrote:
>>> On 7/26/2010 5:30 AM, Rainer Jung wrote:
>>>> I'm a bit undecided whether to port some changes between APR and
>>>> APR-UTIL branches:
>>>>
>>>> - r780882 (wrowe): fix vpath building for xml/expat (removing
>>>> "configure" target in Makefile.in)
>>>>     The change is in the 1.3.x branch, but neither in any older nor newer
>>>> branches.
>>>
>>> I think forward porting is fine, this wasn't much more than an mkdir,
>>> right?
>>
>> You removed running autoconf to create configure:
>>
>> @@ -52,8 +52,6 @@
>>   top_builddir = .
>>
>>
>> -AUTOCONF = autoconf
>> -
>>   INSTALL = @INSTALL@
>>   INSTALL_PROGRAM = @INSTALL_PROGRAM@
>>   INSTALL_DATA = @INSTALL_DATA@
>> @@ -98,9 +96,6 @@
>>                  $(SHELL) configure ; \
>>          fi
>>
>> -configure: configure.in
>> -       $(AUTOCONF)
>> -
>>   config.h: config.h.in config.status
>>          CONFIG_FILES= CONFIG_HEADERS=$(CONFIG_HEADERS) \
>>          $(SHELL) ./config.status
>
> Sorry, I should have read more closely before sending.
>
> autoconf is a packaging tool.  Not a build too.  This didn't belong in Makefile,
> since it's simple enough to have small timing issues when checking out or unpacking
> files from an archive, which triggers such innocuous looking statements.
>
> Forward porting would be terrific;

OK

> does this issue even exist on 0.9 branch though?

Yes:

% cksum apr-util/branches/?.?.x/xml/expat/Makefile.in
285877461       3783    apr-util/branches/0.9.x/xml/expat/Makefile.in
285877461       3783    apr-util/branches/1.0.x/xml/expat/Makefile.in
285877461       3783    apr-util/branches/1.1.x/xml/expat/Makefile.in
285877461       3783    apr-util/branches/1.2.x/xml/expat/Makefile.in
753950660       3724    apr-util/branches/1.3.x/xml/expat/Makefile.in
285877461       3783    apr-util/branches/1.4.x/xml/expat/Makefile.in
285877461       3783    apr-util/branches/1.5.x/xml/expat/Makefile.in

All branches are identical to the version before your change.

Regards,

Rainer

Re: Backporting and forward porting on APR / APR-UTIL

Posted by "William A. Rowe Jr." <wr...@rowe-clan.net>.
On 7/26/2010 10:10 AM, Rainer Jung wrote:
> On 26.07.2010 15:47, William A. Rowe Jr. wrote:
>> On 7/26/2010 5:30 AM, Rainer Jung wrote:
>>> I'm a bit undecided whether to port some changes between APR and
>>> APR-UTIL branches:
>>>
>>> - r780882 (wrowe): fix vpath building for xml/expat (removing
>>> "configure" target in Makefile.in)
>>>    The change is in the 1.3.x branch, but neither in any older nor newer
>>> branches.
>>
>> I think forward porting is fine, this wasn't much more than an mkdir,
>> right?
> 
> You removed running autoconf to create configure:
> 
> @@ -52,8 +52,6 @@
>  top_builddir = .
> 
> 
> -AUTOCONF = autoconf
> -
>  INSTALL = @INSTALL@
>  INSTALL_PROGRAM = @INSTALL_PROGRAM@
>  INSTALL_DATA = @INSTALL_DATA@
> @@ -98,9 +96,6 @@
>                 $(SHELL) configure ; \
>         fi
> 
> -configure: configure.in
> -       $(AUTOCONF)
> -
>  config.h: config.h.in config.status
>         CONFIG_FILES= CONFIG_HEADERS=$(CONFIG_HEADERS) \
>         $(SHELL) ./config.status

Sorry, I should have read more closely before sending.

autoconf is a packaging tool.  Not a build too.  This didn't belong in Makefile,
since it's simple enough to have small timing issues when checking out or unpacking
files from an archive, which triggers such innocuous looking statements.

Forward porting would be terrific; does this issue even exist on 0.9 branch though?

Re: Backporting and forward porting on APR / APR-UTIL

Posted by Rainer Jung <ra...@kippdata.de>.
On 26.07.2010 15:47, William A. Rowe Jr. wrote:
> On 7/26/2010 5:30 AM, Rainer Jung wrote:
>> I'm a bit undecided whether to port some changes between APR and
>> APR-UTIL branches:
>>
>> - r780882 (wrowe): fix vpath building for xml/expat (removing
>> "configure" target in Makefile.in)
>>    The change is in the 1.3.x branch, but neither in any older nor newer
>> branches.
>
> I think forward porting is fine, this wasn't much more than an mkdir, right?

You removed running autoconf to create configure:

@@ -52,8 +52,6 @@
  top_builddir = .


-AUTOCONF = autoconf
-
  INSTALL = @INSTALL@
  INSTALL_PROGRAM = @INSTALL_PROGRAM@
  INSTALL_DATA = @INSTALL_DATA@
@@ -98,9 +96,6 @@
                 $(SHELL) configure ; \
         fi

-configure: configure.in
-       $(AUTOCONF)
-
  config.h: config.h.in config.status
         CONFIG_FILES= CONFIG_HEADERS=$(CONFIG_HEADERS) \
         $(SHELL) ./config.status

>> - r979102 (rjung): Updating buildconf.sh and configure.in for in
>> xml/expat to allow building with libtool 2. Applied to 1.5, 1.4 and 1.3.
>> Should I backport to older branches, or do we want to stay very
>> conservatve here? I'd say backport, because otherwise we might run into
>> the same problem we had with the first tarball for 2.2.16 again, namely
>> that the machine of the RM is to modern, even when preparing e.g. an
>> httpd 2.0 release.
>
> 1.2 isn't shipping or supported.  For luddites, patching 0.9 branch is fine,
> though there isn't likely to be a release except for a security flaw fix.

OK

>> - r979109 (rjung): Updating config.guess and config.sub to allow for
>> better detection of modern platforms. Applied to APR 1.3, ..., trunk and
>> apr-util 1.3, ..., 1.5. I suggest applying to the old branches as well.
>
> Again, the only old branch is 0.9.

OK

Re: Backporting and forward porting on APR / APR-UTIL

Posted by "William A. Rowe Jr." <wr...@rowe-clan.net>.
On 7/26/2010 5:30 AM, Rainer Jung wrote:
> I'm a bit undecided whether to port some changes between APR and
> APR-UTIL branches:
> 
> - r780882 (wrowe): fix vpath building for xml/expat (removing
> "configure" target in Makefile.in)
>   The change is in the 1.3.x branch, but neither in any older nor newer
> branches.

I think forward porting is fine, this wasn't much more than an mkdir, right?

> - r979102 (rjung): Updating buildconf.sh and configure.in for in
> xml/expat to allow building with libtool 2. Applied to 1.5, 1.4 and 1.3.
> Should I backport to older branches, or do we want to stay very
> conservatve here? I'd say backport, because otherwise we might run into
> the same problem we had with the first tarball for 2.2.16 again, namely
> that the machine of the RM is to modern, even when preparing e.g. an
> httpd 2.0 release.

1.2 isn't shipping or supported.  For luddites, patching 0.9 branch is fine,
though there isn't likely to be a release except for a security flaw fix.

> - r979109 (rjung): Updating config.guess and config.sub to allow for
> better detection of modern platforms. Applied to APR 1.3, ..., trunk and
> apr-util 1.3, ..., 1.5. I suggest applying to the old branches as well.

Again, the only old branch is 0.9.