You are viewing a plain text version of this content. The canonical link for it is here.
Posted to commits@apr.apache.org by co...@apache.org on 2009/12/09 02:27:50 UTC
svn commit: r888669 - /apr/apr/trunk/test/testdso.c
Author: covener
Date: Wed Dec 9 01:27:49 2009
New Revision: 888669
URL: http://svn.apache.org/viewvc?rev=888669&view=rev
Log:
let HP-UX on IA64 fall through to the default unix case, using .so instead of .sl.
A quick scan of a couple years of dev@apr only turned up one person reporting
results on this platform, and they had the 8/9 testdso failures as well.
Modified:
apr/apr/trunk/test/testdso.c
Modified: apr/apr/trunk/test/testdso.c
URL: http://svn.apache.org/viewvc/apr/apr/trunk/test/testdso.c?rev=888669&r1=888668&r2=888669&view=diff
==============================================================================
--- apr/apr/trunk/test/testdso.c (original)
+++ apr/apr/trunk/test/testdso.c Wed Dec 9 01:27:49 2009
@@ -38,7 +38,7 @@
#elif defined(DARWIN)
# define MOD_NAME ".libs/mod_test.so"
# define LIB_NAME ".libs/libmod_test.dylib"
-#elif defined(__hpux__) || defined(__hpux)
+#elif (defined(__hpux__) || defined(__hpux)) && !defined(__ia64)
# define MOD_NAME ".libs/mod_test.sl"
# define LIB_NAME ".libs/libmod_test.sl"
#elif defined(_AIX) || defined(__bsdi__)
Re: svn commit: r888669 - /apr/apr/trunk/test/testdso.c
Posted by "William A. Rowe Jr." <wr...@rowe-clan.net>.
Eric Covener wrote:
>
> There is one red herring (starbase.c.h) , and a deprecated facility
> for determining the HW model (model.h) but I think __ia64 is the way
> to go.
sounds good.
Re: svn commit: r888669 - /apr/apr/trunk/test/testdso.c
Posted by Eric Covener <co...@gmail.com>.
On Wed, Dec 9, 2009 at 2:30 PM, William A. Rowe Jr. <wr...@rowe-clan.net> wrote:
> Eric Covener wrote:
>>
>> .sl vs .so does not seem to be one of the behaviors where the PA64
>> behavior matches the Itanium behavior, confirmed with a PA64 APR build
>> and below:
>>
>> http://docs.hp.com/en/B2355-90968/linkerdifferencesiapa.htm
>>
>> Do you object to backporting the testdso.c update in trunk? The test
>> has to predict what libtool will spit out, and this gets HP/IA64 "make
>> check" running clean.
>
> It appears we can count on .sl on PA-32 and PA-64... I was confusing this
> for the availability of -ldl on PA-64 and not PA-32.
>
> Do you want to test for IA, or the absense of PA? I'd look for PA instead,
> as who knows what the next architecture is.
>
Not having any luck with that, browsing their headers it seems to be
all __ia64 or !__ia64.
There is one red herring (starbase.c.h) , and a deprecated facility
for determining the HW model (model.h) but I think __ia64 is the way
to go.
--
Eric Covener
covener@gmail.com
Re: svn commit: r888669 - /apr/apr/trunk/test/testdso.c
Posted by "William A. Rowe Jr." <wr...@rowe-clan.net>.
Eric Covener wrote:
>
> .sl vs .so does not seem to be one of the behaviors where the PA64
> behavior matches the Itanium behavior, confirmed with a PA64 APR build
> and below:
>
> http://docs.hp.com/en/B2355-90968/linkerdifferencesiapa.htm
>
> Do you object to backporting the testdso.c update in trunk? The test
> has to predict what libtool will spit out, and this gets HP/IA64 "make
> check" running clean.
It appears we can count on .sl on PA-32 and PA-64... I was confusing this
for the availability of -ldl on PA-64 and not PA-32.
Do you want to test for IA, or the absense of PA? I'd look for PA instead,
as who knows what the next architecture is.
Re: svn commit: r888669 - /apr/apr/trunk/test/testdso.c
Posted by Eric Covener <co...@gmail.com>.
On Wed, Dec 9, 2009 at 6:53 AM, Eric Covener <co...@gmail.com> wrote:
> On Tue, Dec 8, 2009 at 9:37 PM, William A. Rowe Jr. <wr...@rowe-clan.net> wrote:
>> covener@apache.org wrote:
>>> Author: covener
>>> Date: Wed Dec 9 01:27:49 2009
>>> New Revision: 888669
>>>
>>> URL: http://svn.apache.org/viewvc?rev=888669&view=rev
>>> Log:
>>> let HP-UX on IA64 fall through to the default unix case, using .so instead of .sl.
>>>
>>> A quick scan of a couple years of dev@apr only turned up one person reporting
>>> results on this platform, and they had the 8/9 testdso failures as well.
>>
>> There is a better fix for this, and the test needs to be the same for dso.c,
>> testdso.c and in ./configure.
>>
>> [I'm not really happy that apr doesn't answer such questions for us programatically.]
>>
>> The .sl format is only used for hpux parisc 32 bit. So IA64 is not the only exception.
.sl vs .so does not seem to be one of the behaviors where the PA64
behavior matches the Itanium behavior, confirmed with a PA64 APR build
and below:
http://docs.hp.com/en/B2355-90968/linkerdifferencesiapa.htm
Do you object to backporting the testdso.c update in trunk? The test
has to predict what libtool will spit out, and this gets HP/IA64 "make
check" running clean.
--
Eric Covener
covener@gmail.com
Re: svn commit: r888669 - /apr/apr/trunk/test/testdso.c
Posted by Eric Covener <co...@gmail.com>.
On Tue, Dec 8, 2009 at 9:37 PM, William A. Rowe Jr. <wr...@rowe-clan.net> wrote:
> covener@apache.org wrote:
>> Author: covener
>> Date: Wed Dec 9 01:27:49 2009
>> New Revision: 888669
>>
>> URL: http://svn.apache.org/viewvc?rev=888669&view=rev
>> Log:
>> let HP-UX on IA64 fall through to the default unix case, using .so instead of .sl.
>>
>> A quick scan of a couple years of dev@apr only turned up one person reporting
>> results on this platform, and they had the 8/9 testdso failures as well.
>
> There is a better fix for this, and the test needs to be the same for dso.c,
> testdso.c and in ./configure.
>
> [I'm not really happy that apr doesn't answer such questions for us programatically.]
>
> The .sl format is only used for hpux parisc 32 bit. So IA64 is not the only exception.
>
We currently pick SHL for 32-bit builds on IA64, but if we flipped
over to DLFCN it sounds like it would line up with the .sl/.so
decision. Am I reading too much into the relationship between the
favored API and the shared lib extension?
*-hpux[[1-9]]\.*|*-hpux1[[01]]*)
# shl is specific to hpux(?), and is suboptimal for 64 bit builds,
# and most unlikely to be the choice of 12.x developers.
AC_CHECK_LIB(dld, shl_load, [have_shl=1])
if test "$ac_cv_sizeof_voidp$have_shl" = "41"; then
dsotype=shl; APR_ADDTO(LIBS,-ldld)
fi;;
--
Eric Covener
covener@gmail.com
Re: svn commit: r888669 - /apr/apr/trunk/test/testdso.c
Posted by "William A. Rowe Jr." <wr...@rowe-clan.net>.
covener@apache.org wrote:
> Author: covener
> Date: Wed Dec 9 01:27:49 2009
> New Revision: 888669
>
> URL: http://svn.apache.org/viewvc?rev=888669&view=rev
> Log:
> let HP-UX on IA64 fall through to the default unix case, using .so instead of .sl.
>
> A quick scan of a couple years of dev@apr only turned up one person reporting
> results on this platform, and they had the 8/9 testdso failures as well.
There is a better fix for this, and the test needs to be the same for dso.c,
testdso.c and in ./configure.
[I'm not really happy that apr doesn't answer such questions for us programatically.]
The .sl format is only used for hpux parisc 32 bit. So IA64 is not the only exception.