httpd-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Ruediger Pluem <rpl...@apache.org>
Subject Re: Preferred versions of libtool and autoconf for T&R
Date Sat, 29 Nov 2008 21:47:45 GMT


On 11/29/2008 08:17 PM, Arfrever Frehtes Taifersar Arahesis wrote:
> 2008-11-29 15:49:36 Plüm, Rüdiger, VF-Group napisał(a):
>>> -----Urspr�ngliche Nachricht-----
>>> Von: "Pl�m, R�diger, VF-Group" ] 
>>> Gesendet: Samstag, 29. November 2008 15:19
>>> An: dev@httpd.apache.org
>>> Betreff: Re: Preferred versions of libtool and autoconf for T&R
>>>
>>>  
>>>
>>>> -----Urspr�ngliche Nachricht-----
>>>> Von: William A. Rowe, Jr.  
>>>> Gesendet: Samstag, 29. November 2008 04:06
>>>> An: dev@httpd.apache.org
>>>> Betreff: Re: Preferred versions of libtool and autoconf for T&R
>>>>
>>>> Ruediger Pluem wrote:
>>>>> Any objections going with autoconf 2.63 and libtool 1.5.26?
>>>>>
>>>>> If autoconf 2.63 is seen as too risky I would go back to 
>>>> autoconf 2.61.
>>>>
>>>> I see no remaining issues for 2.63... solid choice.  The endianess
>>>> issues of 2.62 should all be addressed.
>>>>
>>> autoconf 2.63 still emits the warnings for APR / APR-UTIL configure
>>> options passed to it. The following patch should fix this:
>>>
>>> Index: configure.in
>>> ===================================================================
>>> --- configure.in        (revision 721659)
>>> +++ configure.in        (working copy)
>>> @@ -18,6 +18,10 @@
>>>  sinclude(build/find_apu.m4)
>>>  sinclude(acinclude.m4)
>>>
>>> +dnl We don't want our confingure to emit any warnings for any
>>> +dnl APR / APR-UTIL configure options
> 
> Why do you want to pass invalid options to configure?
> (Also: s/confingure/configure/)


I hope the comments to my patch below will explain this.

> 
>>> +AC_DISABLE_OPTION_CHECKING
>>> +
>>>  dnl XXX we can't just use AC_PREFIX_DEFAULT because that 
>>> isn't subbed in
>>>  dnl by configure until it is too late.  Is that how it 
>>> should be or not?
>>>  dnl Something seems broken here.
>>>
>>> Any objections?
>>>
>>> Otherwise I would commit to trunk and propose it for backport.
>> This does not work with older autoconf versions :-(.
>>
>> Anyone an idea for a code that only calls
>>
>> AC_DISABLE_OPTION_CHECKING
>>
>> if it is defined?
> 
> ifdef([AC_DISABLE_OPTION_CHECKING], [AC_DISABLE_OPTION_CHECKING])

That did the trick. Thanks. In my first test I missed the [] around the
first parameter.

So I propose the following patch for trunk which worked fine for me with
autoconf 2.60 from SuSE 10.2 and vanilla autoconf 2.63:

Index: configure.in
===================================================================
--- configure.in        (Revision 721717)
+++ configure.in        (Arbeitskopie)
@@ -18,6 +18,15 @@
 sinclude(build/find_apu.m4)
 sinclude(acinclude.m4)

+dnl Later versions of autoconf (>= 2.62) by default cause the produced
+dnl configure script to emit at least warnings when it comes across unknown
+dnl command line options. These versions also have the macro
+dnl AC_DISABLE_OPTION_CHECKING defined which turns this off by default.
+dnl We want to have this turned off here since our configure calls can
+dnl contain options for APR / APR-UTIL configure that are unkown to us.
+dnl So avoid confusing the user by turning this off. See also PR 45221.
+ifdef([AC_DISABLE_OPTION_CHECKING], [AC_DISABLE_OPTION_CHECKING])
+
 dnl XXX we can't just use AC_PREFIX_DEFAULT because that isn't subbed in
 dnl by configure until it is too late.  Is that how it should be or not?
 dnl Something seems broken here.


Any objections?

As soon as this is in trunk I would propose it for backport and use
autoconf 2.63 for T&R if this gets backported or should I stay with
2.61 and we try 2.63 for the next T&R?

Regards

Rüdiger


Mime
View raw message