Re: svn commit: r816074 - in /httpd/mod_ftp/trunk/modules/ftp: ftp_commands.c ftp_internal.h ftp_util.c

2009-09-17 Thread Jim Jagielski


On Sep 17, 2009, at 4:18 AM, William A. Rowe, Jr. wrote:


rj...@apache.org wrote:


Modified: httpd/mod_ftp/trunk/modules/ftp/ftp_internal.h
URL: 
http://svn.apache.org/viewvc/httpd/mod_ftp/trunk/modules/ftp/ftp_internal.h?rev=816074&r1=816073&r2=816074&view=diff
=
=
=
=
=
=
=
=
=
=
--- httpd/mod_ftp/trunk/modules/ftp/ftp_internal.h (original)
+++ httpd/mod_ftp/trunk/modules/ftp/ftp_internal.h Thu Sep 17  
07:00:24 2009

@@ -84,6 +84,8 @@

#if APR_HAVE_SYS_STAT_H
#include 
+#elif HAVE_SYS_STAT_H
+#include 
#endif


NAK.  The fragment above makes zero sense, please revert.

#if APR_HAVE_SYS_STAT_H

should be sufficent.  If apr does not provide it consistently, and  
httpd

has, then the test becomes

#ifdef HAVE_SYS_STAT_H

which is an altogether different beast.  First clue that the code  
above was
wrong is that you included the same code for both cases.  So  
provided that
you had no clue if APR consistently provided this and you wanted to  
rely

upon config.h, then it becomes

#if (defined(APR_HAVE_SYS_STAT_H) && APR_HAVE_SYS_STAT_H) \
   || defined(HAVE_SYS_STAT_H)
#include...

but we know we don't need to go that far.



Grok ./modules/generators/mod_cgid.c

  /* ### should be tossed in favor of APR */
  #include 

btw :)

I'm sure there's some history there...


Re: svn commit: r816074 - in /httpd/mod_ftp/trunk/modules/ftp: ftp_commands.c ftp_internal.h ftp_util.c

2009-09-17 Thread Rainer Jung
On 17.09.2009 10:18, William A. Rowe, Jr. wrote:
> rj...@apache.org wrote:
>>
>> Modified: httpd/mod_ftp/trunk/modules/ftp/ftp_internal.h
>> URL: 
>> http://svn.apache.org/viewvc/httpd/mod_ftp/trunk/modules/ftp/ftp_internal.h?rev=816074&r1=816073&r2=816074&view=diff
>> ==
>> --- httpd/mod_ftp/trunk/modules/ftp/ftp_internal.h (original)
>> +++ httpd/mod_ftp/trunk/modules/ftp/ftp_internal.h Thu Sep 17 07:00:24 2009
>> @@ -84,6 +84,8 @@
>>  
>>  #if APR_HAVE_SYS_STAT_H
>>  #include 
>> +#elif HAVE_SYS_STAT_H
>> +#include 
>>  #endif
> 
> NAK.  The fragment above makes zero sense, please revert.
> 
> #if APR_HAVE_SYS_STAT_H
> 
> should be sufficent.  If apr does not provide it consistently, and httpd
> has, then the test becomes
> 
> #ifdef HAVE_SYS_STAT_H
> 
> which is an altogether different beast.

Sorry, of course.

> First clue that the code above was
> wrong is that you included the same code for both cases.

This I don't understand. Which cases?

> So provided that
> you had no clue if APR consistently provided this and you wanted to rely

I did have a clue (at least I thought so ;) ), I posted to d...@apr
separately. It does *not* provide it.

> upon config.h, then it becomes
> 
> #if (defined(APR_HAVE_SYS_STAT_H) && APR_HAVE_SYS_STAT_H) \
> || defined(HAVE_SYS_STAT_H)
> #include...
> 
> but we know we don't need to go that far.

Hmmm? Why doesn't that look like the right thing? You want to use only
HAVE_SYS_STAT_H instead? I thought allowing the APR variant made some
sense to support a forthcoming APR version, that might have the APR_
variant defined.

> I'll answer the question you raised on list tomorrow, although it certainly
> seemed like an apr question and not an httpd question.

Yes, see d...@apr.

Not that after reverting the source doesn't build on Solaris using gcc
4.1. The undefined file permission symbols make it err.

Regards,

Rainer


Re: svn commit: r816074 - in /httpd/mod_ftp/trunk/modules/ftp: ftp_commands.c ftp_internal.h ftp_util.c

2009-09-17 Thread William A. Rowe, Jr.
rj...@apache.org wrote:
> 
> Modified: httpd/mod_ftp/trunk/modules/ftp/ftp_internal.h
> URL: 
> http://svn.apache.org/viewvc/httpd/mod_ftp/trunk/modules/ftp/ftp_internal.h?rev=816074&r1=816073&r2=816074&view=diff
> ==
> --- httpd/mod_ftp/trunk/modules/ftp/ftp_internal.h (original)
> +++ httpd/mod_ftp/trunk/modules/ftp/ftp_internal.h Thu Sep 17 07:00:24 2009
> @@ -84,6 +84,8 @@
>  
>  #if APR_HAVE_SYS_STAT_H
>  #include 
> +#elif HAVE_SYS_STAT_H
> +#include 
>  #endif

NAK.  The fragment above makes zero sense, please revert.

#if APR_HAVE_SYS_STAT_H

should be sufficent.  If apr does not provide it consistently, and httpd
has, then the test becomes

#ifdef HAVE_SYS_STAT_H

which is an altogether different beast.  First clue that the code above was
wrong is that you included the same code for both cases.  So provided that
you had no clue if APR consistently provided this and you wanted to rely
upon config.h, then it becomes

#if (defined(APR_HAVE_SYS_STAT_H) && APR_HAVE_SYS_STAT_H) \
|| defined(HAVE_SYS_STAT_H)
#include...

but we know we don't need to go that far.

I'll answer the question you raised on list tomorrow, although it certainly
seemed like an apr question and not an httpd question.