Groups | Search | Server Info | Keyboard shortcuts | Login | Register [http] [https] [nntp] [nntps]

Groups > comp.programming > #2983 > unrolled thread

Should the passed command parameters result into the same error messages for a configuration test?

Back to article view | Back to comp.programming

Contents

  Should the passed command parameters result into the same error messages for a configuration test? Markus Elfring <Markus.Elfring@web.de> - 2013-02-09 12:15 +0100
    Re: Should the passed command parameters result into the same error messages for a configuration test? Rainer Weikusat <rweikusat@mssgmbh.com> - 2013-02-10 18:27 +0000
      Re: Should the passed command parameters result into the same error messages for a configuration test? Markus Elfring <Markus.Elfring@web.de> - 2013-02-11 09:55 +0100
        Re: Should the passed command parameters result into the same error messages for a configuration test? Johann Klammer <klammerj@NOSPAM.a1.net> - 2013-02-11 13:38 +0100
          Re: Should the passed command parameters result into the same error messages for a configuration test? Markus Elfring <Markus.Elfring@web.de> - 2013-02-11 18:28 +0100
            Re: Should the passed command parameters result into the same error messages for a configuration test? Johann Klammer <klammerj@NOSPAM.a1.net> - 2013-02-11 19:35 +0100
              Re: Should the passed command parameters result into the same error   messages for a configuration test? Markus Elfring <Markus.Elfring@web.de> - 2013-02-11 20:02 +0100
        Re: Should the passed command parameters result into the same error messages for a configuration test? Rainer Weikusat <rweikusat@mssgmbh.com> - 2013-02-13 11:59 +0000
          Re: Should the passed command parameters result into the same error messages for a configuration test? Markus Elfring <Markus.Elfring@web.de> - 2013-02-13 20:57 +0100

#2983 — Should the passed command parameters result into the same error messages for a configuration test?

Hello,

I try to generate the Luyten release of a library combination with the software
tool "g++ 4.7.2" on my openSUSE Tumbleweed system.

I wonder about the following results here.
http://gcc.gnu.org/ml/gcc-help/2013-02/msg00032.html

elfring@Sonne:~/Projekte/OpalVoip/OPAL/3.10.9> ./configure --disable-libavcodec
...
checking PTLIB has expat... yes
checking PTLIB has video... no
...

Excerpt from ~/Projekte/OpalVoip/OPAL/3.10.9/config.log:
...
/usr/local/lib64/libpt_s.a(vsdl.o): In function `_GLOBAL__sub_I_vsdl.cxx':
vsdl.cxx:(.text.startup+0x42): undefined reference to
`PPlugin_PVideoInputDevice_Application_link()'
/usr/local/lib64/libpt_s.a(vfakeio.o): In function `_GLOBAL__sub_I_vfakeio.cxx':
vfakeio.cxx:(.text.startup+0x42): undefined reference to
`PPlugin_PVideoInputDevice_Application_link()'
...


elfring@Sonne:~/Projekte/OpalVoip/OPAL/Probe> cat Video1.cpp
#define PACKAGE_NAME ""
#define PACKAGE_TARNAME ""
#define PACKAGE_VERSION ""
#define PACKAGE_STRING ""
#define PACKAGE_BUGREPORT ""
#define PACKAGE_URL ""
#define STDC_HEADERS 1
#define OPAL_MAJOR 3
#define OPAL_MINOR 10
#define OPAL_BUILD 9
#define OPAL_VERSION "3.10.9"
#define OPAL_PLUGIN_DIR "/usr/local/lib64/opal-3.10.9"
#define OPAL_PTLIB_SSL 1
#define OPAL_PTLIB_ASN 1
#define OPAL_PTLIB_EXPAT 1

#include <ptbuildopts.h>
#include <ptlib.h>
#include <ptlib/video.h>
int main(void)
{
 PVideoChannel vc;
 return 0;
}
elfring@Sonne:~/Projekte/OpalVoip/OPAL/Probe> LANG=C g++ -o Video1 -Wall -Wextra
-Wstrict-aliasing -Wfloat-equal -Wno-comment -Wno-unused -Winit-self
-Wno-missing-field-initializers -Wreorder -felide-constructors -fexceptions
-fPIC -DP_64BIT -DPTRACING=1 -D_REENTRANT -D_GNU_SOURCE=1 -I/usr/local/include
-I/usr/include/SDL -L/usr/local/lib64 -lpt_s -lrt -lsasl2 -lldap -llber -lldap_r
-lssl -lcrypto -lexpat -llua -lSDL -lpthread -lodbc -lresolv -ldl Video1.cpp
/tmp/ccc6uWN0.o: In function `main':
Video1.cpp:(.text+0x17): undefined reference to `PVideoChannel::PVideoChannel()'
Video1.cpp:(.text+0x2b): undefined reference to `PVideoChannel::~PVideoChannel()'
/tmp/ccc6uWN0.o: In function `__static_initialization_and_destruction_0(int, int)':
Video1.cpp:(.text+0x8c): undefined reference to
`PFactoryLoader::PluginLoaderStartup_link()'
Video1.cpp:(.text+0x97): undefined reference to
`PPlugin_PVideoInputDevice_FakeVideo_link()'
Video1.cpp:(.text+0xa2): undefined reference to
`PPlugin_PVideoInputDevice_Application_link()'
Video1.cpp:(.text+0xad): undefined reference to
`PPlugin_PVideoInputDevice_FFMPEG_link()'
Video1.cpp:(.text+0xb8): undefined reference to
`PPlugin_PVideoInputDevice_YUVFile_link()'
Video1.cpp:(.text+0xc3): undefined reference to
`PPlugin_PVideoOutputDevice_SDL_link()'
collect2: error: ld returned 1 exit status


I find it also especially strange that the reduced command parameter set
(deletion of duplicated options and preprocessor symbols) shows differences in
the error messages.

I would appreciate your advices.

Regards,
Markus

[toc] | [next] | [standalone]

#2995

Markus Elfring <Markus.Elfring@web.de> writes:
> I try to generate the Luyten release of a library combination with the software
> tool "g++ 4.7.2" on my openSUSE Tumbleweed system.
>
> I wonder about the following results here.
> http://gcc.gnu.org/ml/gcc-help/2013-02/msg00032.html
>
> elfring@Sonne:~/Projekte/OpalVoip/OPAL/3.10.9> ./configure --disable-libavcodec
> ...
> checking PTLIB has expat... yes
> checking PTLIB has video... no
> ...
>
> Excerpt from ~/Projekte/OpalVoip/OPAL/3.10.9/config.log:
> ...
> /usr/local/lib64/libpt_s.a(vsdl.o): In function `_GLOBAL__sub_I_vsdl.cxx':
> vsdl.cxx:(.text.startup+0x42): undefined reference to
> `PPlugin_PVideoInputDevice_Application_link()'
> /usr/local/lib64/libpt_s.a(vfakeio.o): In function `_GLOBAL__sub_I_vfakeio.cxx':
> vfakeio.cxx:(.text.startup+0x42): undefined reference to
> `PPlugin_PVideoInputDevice_Application_link()'

[Video1.cpp]

> elfring@Sonne:~/Projekte/OpalVoip/OPAL/Probe> LANG=C g++ -o Video1 -Wall -Wextra
> -Wstrict-aliasing -Wfloat-equal -Wno-comment -Wno-unused -Winit-self
> -Wno-missing-field-initializers -Wreorder -felide-constructors -fexceptions
> -fPIC -DP_64BIT -DPTRACING=1 -D_REENTRANT -D_GNU_SOURCE=1 -I/usr/local/include
> -I/usr/include/SDL -L/usr/local/lib64 -lpt_s -lrt -lsasl2 -lldap -llber -lldap_r
> -lssl -lcrypto -lexpat -llua -lSDL -lpthread -lodbc -lresolv -ldl Video1.cpp
> /tmp/ccc6uWN0.o: In function `main':
> Video1.cpp:(.text+0x17): undefined reference to `PVideoChannel::PVideoChannel()'
> Video1.cpp:(.text+0x2b): undefined reference to `PVideoChannel::~PVideoChannel()'
> /tmp/ccc6uWN0.o: In function `__static_initialization_and_destruction_0(int, int)':
> Video1.cpp:(.text+0x8c): undefined reference to
> `PFactoryLoader::PluginLoaderStartup_link()'
> Video1.cpp:(.text+0x97): undefined reference to
> `PPlugin_PVideoInputDevice_FakeVideo_link()'
> Video1.cpp:(.text+0xa2): undefined reference to
> `PPlugin_PVideoInputDevice_Application_link()'
> Video1.cpp:(.text+0xad): undefined reference to
> `PPlugin_PVideoInputDevice_FFMPEG_link()'
> Video1.cpp:(.text+0xb8): undefined reference to
> `PPlugin_PVideoInputDevice_YUVFile_link()'
> Video1.cpp:(.text+0xc3): undefined reference to
> `PPlugin_PVideoOutputDevice_SDL_link()'
> collect2: error: ld returned 1 exit status
>
> I find it also especially strange that the reduced command parameter set
> (deletion of duplicated options and preprocessor symbols) shows differences in
> the error messages.
>
> I would appreciate your advices.

Your posting contains to random lines from an unknown config.log and a
g++ command presumably supposed to compile some C++ source code file
and link it in order to create an executable binary from it which
failed because the link editor couldn't find all required symbols.

This leaves an interesting question open, namely "What's your
question?".

[toc] | [prev] | [next] | [standalone]

#2996

> ..., namely "What's your question?".

1. Other software developers generated the affected library successfully.

Examples:
http://pkgs.fedoraproject.org/cgit/opal.git
http://packages.debian.org/experimental/libopal3.10.9
https://build.opensuse.org/package/show?project=openSUSE%3AFactory&package=opal

So I wonder why I stumble on "undefined references" for my build configuration
variant. I expect that all needed symbols should be available also for the
linker. Why can they not be resolved here?


2. The shown configuration test is generated by the Autotools. The command
contains duplicate options and preprocessor symbols. I dared to shorten the call
a bit. Why are the displayed messages so different after my adjustment?

Regards,
Markus

[toc] | [prev] | [next] | [standalone]

#2997

Markus Elfring wrote:
>> ..., namely "What's your question?".
>
> 1. Other software developers generated the affected library successfully.
>
> Examples:
> http://pkgs.fedoraproject.org/cgit/opal.git
> http://packages.debian.org/experimental/libopal3.10.9
> https://build.opensuse.org/package/show?project=openSUSE%3AFactory&package=opal
>
> So I wonder why I stumble on "undefined references" for my build configuration
> variant. I expect that all needed symbols should be available also for the
> linker. Why can they not be resolved here?
>
>
> 2. The shown configuration test is generated by the Autotools. The command
> contains duplicate options and preprocessor symbols. I dared to shorten the call
> a bit. Why are the displayed messages so different after my adjustment?
>
> Regards,
> Markus
>

I believe that there may have been changes to the linker or compiler, so 
the order of arguments given does matter now.

try 'input' 'libraries' 'output'

instead of 'output' 'libraries' 'input'

I believe the reason for doing this, was that linker resolved too many 
symbols when linking, which resulted in symbols being bound that where 
not used by the .o file, resulting in namespace contamination.

This is all just guesswork.
Good luck.

[toc] | [prev] | [next] | [standalone]

#2999

> I believe that there may have been changes to the linker or compiler, so the
> order of arguments given does matter now.

Thanks for your suggestion.

I would like to be sure that the adjusted parameter set would be equivalent to
the Autotools generated one which contains duplicate values. I guess that a
shorter parameter list will be easier to read, won't it?


> This is all just guesswork.

I hope that this situation can be improved.

Regards,
Markus

[toc] | [prev] | [next] | [standalone]

#3000

Markus Elfring wrote:
>> This is all just guesswork.
>
> I hope that this situation can be improved.
>
I meant that _I_ was just guessing at this. It may not be your actual 
problem. I do not even know for sure if those things really have 
changed, as I have merely been reading about this somewhere on the 
net(cannot remember where) but not experienced it myself.

sloppy autotools scripts getting the libraries wrong is also a popular 
thing...

[toc] | [prev] | [next] | [standalone]

#3001 — Re: Should the passed command parameters result into the same error messages for a configuration test?

> sloppy autotools scripts getting the libraries wrong is also a popular thing...

I hope that the following configuration check will also work on my system in the
near future as expected.
http://opalvoip.svn.sourceforge.net/viewvc/opalvoip/opal/tags/v3_12_0/configure.ac?revision=28491&view=markup
:

...
OPAL_CHECK_PTLIB([video], [PTLIB_VIDEO])
...

Regards,
Markus

[toc] | [prev] | [next] | [standalone]

#3011

Markus Elfring <Markus.Elfring@web.de> writes:
>> ..., namely "What's your question?".
>
> 1. Other software developers generated the affected library successfully.
>
> Examples:
> http://pkgs.fedoraproject.org/cgit/opal.git
> http://packages.debian.org/experimental/libopal3.10.9
> https://build.opensuse.org/package/show?project=openSUSE%3AFactory&package=opal
>
> So I wonder why I stumble on "undefined references" for my build configuration
> variant. I expect that all needed symbols should be available also for the
> linker. Why can they not be resolved here?

That's a question nobody except you can answer. But "it worked on
other installations, hence, it must work on this installation, too" is
a non-sequitur: Obviously 'something must be different' on the
installation where it didn't work.

> 2. The shown configuration test is generated by the Autotools. The command
> contains duplicate options and preprocessor symbols. I dared to shorten the call
> a bit. Why are the displayed messages so different after my adjustment?

All you posted is 'a test program' (which might or might not be identical to
the one used during configuration) and 'a compilation command' which
you believe to be semantically equivalent to 'another compilation
command' nobody except you saw. The fact that there was a difference
in behaviour (assuming there was actually a difference in behaviour
and not just a copy&paste error you made which looks as if there was
a difference in behaviour) implies that "both must be different"
despite you don't think they are. But beyond the obviously suggestion
that your opinion regarding the 'equivalence' of 'all of A' and 'all
of B' cannot possibly be true (here including that you actually posted
all relevant output for both cases), there's nothing anyone can say to
that.

BTW: Please stop this 'posted & mailed' nonsense --- I don't mind
getting mail in reply to USENET postings, that's why I use a real
e-mail address, but a mail in addition to a posting is somewhat
useless.

[toc] | [prev] | [next] | [standalone]

#3018

> But beyond the obviously suggestion that your opinion regarding the 'equivalence'
> of 'all of A' and 'all of B' cannot possibly be true (here including that you
> actually posted all relevant output for both cases), there's nothing anyone
> can say to that.

I imagine that this situation can change if more software developers and 
testers would dare to try out the shown use case.

Regards,
Markus

[toc] | [prev] | [standalone]

Back to top | Article view | comp.programming

csiph-web