subject:"\[cmake\-developers\] GenerateExportHeader macro in CMake\?"

Re: [cmake-developers] GenerateExportHeader macro in CMake?

2011-08-25 Thread Stephen


On 08/24/2011 08:49 PM, David Cole wrote:

On Wed, Aug 24, 2011 at 2:27 PM, Stephen Kellysteve...@gmail.com  wrote:

David Cole wrote:


The test itself passes on AIX, as seen here:
http://www.cdash.org/CDash/testDetails.php?test=109727151build=1460504

It's only the particular machine that you pointed to that has
problems. There are *several* tests that fail on that machine with the
space in the path. That's one of the reasons it's only in Nightly
and not Nightly Expected


Even after my change this one is still failing.

Someone with access to the box would have to
cd ${CMAKE_BUILD_DIR}/Tests/Module/GenerateExportHeader/libsharedtest/fail1
rm -rf *
cmake ../test1

then inspect the trivial ../test1/CMakeLists.txt to see why it is failing.

I made another guess and pushed another patch, but it's a quite 
inefficient way to fix this issue. Does anyone have access to that box?



You're right, it should be CMAKE_CXX_COMPILER_ID (or _C_). And do you
have the right logical sense of the if test? Do you want that simple
main for the Watcom compiler, or should it be a NOT? (I'm just
asking to double-check, I haven't dug in and read enough of your test
to fully appreciate the chunk shown in the gitweb diff view.)


Yes. CDash seems to fail if there is no executable of the correct name 
which runs successfully.


For Watcom I create a trivial executable with the appropriate name 
before I return(). For not watcom the tests are run and the trivial 
executable is not created. I think it's correct as-is. The machine 
hasn't yet finished to confirm that though.


Thanks,

Steve.
___
cmake-developers mailing list
cmake-developers@cmake.org
http://public.kitware.com/cgi-bin/mailman/listinfo/cmake-developers

Re: [cmake-developers] GenerateExportHeader macro in CMake?

2011-08-25 Thread David Cole

On Thu, Aug 25, 2011 at 12:35 PM, Stephen steve...@gmail.com wrote:
 On 08/24/2011 08:49 PM, David Cole wrote:

 On Wed, Aug 24, 2011 at 2:27 PM, Stephen Kellysteve...@gmail.com  wrote:

 David Cole wrote:

 The test itself passes on AIX, as seen here:
 http://www.cdash.org/CDash/testDetails.php?test=109727151build=1460504

 It's only the particular machine that you pointed to that has
 problems. There are *several* tests that fail on that machine with the
 space in the path. That's one of the reasons it's only in Nightly
 and not Nightly Expected

 Even after my change this one is still failing.

 Someone with access to the box would have to
 cd ${CMAKE_BUILD_DIR}/Tests/Module/GenerateExportHeader/libsharedtest/fail1
 rm -rf *
 cmake ../test1

 then inspect the trivial ../test1/CMakeLists.txt to see why it is failing.

 I made another guess and pushed another patch, but it's a quite inefficient
 way to fix this issue. Does anyone have access to that box?


Please DO NOT push any more changes just to address this machine. As I
said yesterday, there are several tests (25 today) that fail on that
machine and we are not basing any of our decisions on that machine's
CDash submissions.

  http://www.cdash.org/CDash/viewTest.php?onlyfailedbuildid=1465110

If we get to the point where all the other tests are passing there,
and yours is the only failing test, then at that point, I'll come back
to you for more. That's unlikely to happen soon, as we have a
demonstrably successful submission from another AIX machine in the
Nightly Expected section.


 You're right, it should be CMAKE_CXX_COMPILER_ID (or _C_). And do you
 have the right logical sense of the if test? Do you want that simple
 main for the Watcom compiler, or should it be a NOT? (I'm just
 asking to double-check, I haven't dug in and read enough of your test
 to fully appreciate the chunk shown in the gitweb diff view.)

 Yes. CDash seems to fail if there is no executable of the correct name which
 runs successfully.

 For Watcom I create a trivial executable with the appropriate name before I
 return(). For not watcom the tests are run and the trivial executable is not
 created. I think it's correct as-is. The machine hasn't yet finished to
 confirm that though.

 Thanks,

 Steve.

___
cmake-developers mailing list
cmake-developers@cmake.org
http://public.kitware.com/cgi-bin/mailman/listinfo/cmake-developers

Re: [cmake-developers] GenerateExportHeader macro in CMake?

2011-08-24 Thread Stephen Kelly

Stephen Kelly wrote:

 Yep, I've just had a look and possibily fixed the Intel and HPUX cases.
 

I did indeed fix the Intel and HPUX cases.

 The AIX one fails because the visibility test passes, because
 check_cxx_compiler_flag does not have a FAIL_REGEX for its warning:
 
 http://www.cdash.org/CDash/testDetails.php?test=109778717build=1458523
 
 http://www.cdash.org/CDash/testDetails.php?test=110283056build=1458523
 
 I've also pushed a possible fix for that to check_cxx_compiler_flag on the
 branch.

The FAIL_REGEX I added for AIX was successful. However, now that box reports 
some garbage:

  ld: 0706-014 The -b
  Tests/CMake-
build/Tests/Module/GenerateExportHeader/libsharedtest/fail1/libshared
  option is not recognized.

Someone with access to the box would have to find out what's going wrong. It 
might have been caused by a space in the path, which I've just added a fix 
for. 

It might be worth running some tests twice: Once from a path with a space in 
it, and once without. Or run all tests from a path with a space in it to 
prevent these kinds of bugs.

I would prefer to turn off the tests for it anyway, but I don't know what 
platform test to make. What is the equivalent to 

if (WIN32)
  return()
endif()

for AIX?

 The CentOS machine seems to have problems unrelated to this test:
 
 http://www.cdash.org/CDash/testDetails.php?test=109887854build=1457339
 
 http://www.cdash.org/CDash/viewTest.php?buildid=1457339
 
 All have ' undefined reference to `_Unwind_Resume'' in their output.
 

Do I have to do anything about this?

 That only leaves Watcom as you say, and I've just tried to simplify the
 tests for that compiler. If the timeouts persist, I can start just
 disabling some tests for that compiler.
 

This one is still timing out. I've disable the test for it with 
(CMAKE_COMPILER_FLAG MATCHES Watcom).

Thanks,

Steve.


___
cmake-developers mailing list
cmake-developers@cmake.org
http://public.kitware.com/cgi-bin/mailman/listinfo/cmake-developers

Re: [cmake-developers] GenerateExportHeader macro in CMake?

2011-08-24 Thread David Cole

On Wed, Aug 24, 2011 at 2:02 PM, Stephen Kelly steve...@gmail.com wrote:
Stephen Kelly wrote:

Yep, I've just had a look and possibily fixed the Intel and HPUX cases.

I did indeed fix the Intel and HPUX cases.

Thank you.

The AIX one fails because the visibility test passes, because
check_cxx_compiler_flag does not have a FAIL_REGEX for its warning:

http://www.cdash.org/CDash/testDetails.php?test=109778717build=1458523

http://www.cdash.org/CDash/testDetails.php?test=110283056build=1458523

I've also pushed a possible fix for that to check_cxx_compiler_flag on the
branch.

The FAIL_REGEX I added for AIX was successful.

That's fine...

However, now that box reports
some garbage:

ld: 0706-014 The -b
Tests/CMake-
build/Tests/Module/GenerateExportHeader/libsharedtest/fail1/libshared
option is not recognized.

Someone with access to the box would have to find out what's going wrong. It
might have been caused by a space in the path, which I've just added a fix
for.

It might be worth running some tests twice: Once from a path with a space in
it, and once without. Or run all tests from a path with a space in it to
prevent these kinds of bugs.

I would prefer to turn off the tests for it anyway, but I don't know what
platform test to make. What is the equivalent to

if (WIN32)
return()
endif()

for AIX?

The test itself passes on AIX, as seen here:
http://www.cdash.org/CDash/testDetails.php?test=109727151build=1460504

It's only the particular machine that you pointed to that has
problems. There are *several* tests that fail on that machine with the
space in the path. That's one of the reasons it's only in Nightly
and not Nightly Expected

The CentOS machine seems to have problems unrelated to this test:

http://www.cdash.org/CDash/testDetails.php?test=109887854build=1457339

http://www.cdash.org/CDash/viewTest.php?buildid=1457339

All have ' undefined reference to `_Unwind_Resume'' in their output.

Do I have to do anything about this?

No -- this will go away eventually. Either by being fixed or being banished.

That only leaves Watcom as you say, and I've just tried to simplify the
tests for that compiler. If the timeouts persist, I can start just
disabling some tests for that compiler.

This one is still timing out. I've disable the test for it with
(CMAKE_COMPILER_FLAG MATCHES Watcom).

You mean CMAKE_COMPILER_ID as in this commit, right?
http://cmake.org/gitweb?p=cmake.git;a=commitdiff;h=99b2aabd460d0c508ffa7b21283a0512e322e717

That's fine.

Thanks,

Steve.

Thank *you* -- your persistence has paid off.

I think tomorrow it will be passing enough to merge over to 'master' ... :-)

David
___
cmake-developers mailing list
cmake-developers@cmake.org
http://public.kitware.com/cgi-bin/mailman/listinfo/cmake-developers

Re: [cmake-developers] GenerateExportHeader macro in CMake?

2011-08-24 Thread Stephen Kelly

David Cole wrote:

 The test itself passes on AIX, as seen here:
 http://www.cdash.org/CDash/testDetails.php?test=109727151build=1460504
 
 It's only the particular machine that you pointed to that has
 problems. There are *several* tests that fail on that machine with the
 space in the path. That's one of the reasons it's only in Nightly
 and not Nightly Expected

Ok. Nevertheless I think it would make sense for cdash to always check out 
and build in a directory with a space in it. There could be other bugs 
lurking.

 
 That only leaves Watcom as you say, and I've just tried to simplify the
 tests for that compiler. If the timeouts persist, I can start just
 disabling some tests for that compiler.


 This one is still timing out. I've disable the test for it with
 (CMAKE_COMPILER_FLAG MATCHES Watcom).

 
 You mean CMAKE_COMPILER_ID as in this commit, right?
 
http://cmake.org/gitweb?p=cmake.git;a=commitdiff;h=99b2aabd460d0c508ffa7b21283a0512e322e717
 
 That's fine.
 

On closer inspection I figured it should be CMAKE_CXX_COMPILER_ID. I don't 
know if CMAKE_COMPILER_ID would work too...

 
 Thank *you* -- your persistence has paid off.
 
 I think tomorrow it will be passing enough to merge over to 'master' ...
 :-)
 

Great. Hopefully it will be useful to many.

Steve.


___
cmake-developers mailing list
cmake-developers@cmake.org
http://public.kitware.com/cgi-bin/mailman/listinfo/cmake-developers

Re: [cmake-developers] GenerateExportHeader macro in CMake?

2011-08-24 Thread David Cole

On Wed, Aug 24, 2011 at 2:27 PM, Stephen Kelly steve...@gmail.com wrote:
David Cole wrote:

The test itself passes on AIX, as seen here:
http://www.cdash.org/CDash/testDetails.php?test=109727151build=1460504

Ok. Nevertheless I think it would make sense for cdash to always check out
and build in a directory with a space in it. There could be other bugs
lurking.

The vast majority of our dashboard builds are done on build trees and
source trees that have a space in the full path name. We only do
non-space-in-the-path builds on machines where there are problems with
other tools over which we have no control.

We all strive to have cmake, ctest and cpack work flawlessly when
there are spaces in file and path names. We only punt on a given
dashboard machine when we have to.

That only leaves Watcom as you say, and I've just tried to simplify the
tests for that compiler. If the timeouts persist, I can start just
disabling some tests for that compiler.

This one is still timing out. I've disable the test for it with
(CMAKE_COMPILER_FLAG MATCHES Watcom).

You mean CMAKE_COMPILER_ID as in this commit, right?

http://cmake.org/gitweb?p=cmake.git;a=commitdiff;h=99b2aabd460d0c508ffa7b21283a0512e322e717

That's fine.

On closer inspection I figured it should be CMAKE_CXX_COMPILER_ID. I don't
know if CMAKE_COMPILER_ID would work too...

You're right, it should be CMAKE_CXX_COMPILER_ID (or _C_). And do you
have the right logical sense of the if test? Do you want that simple
main for the Watcom compiler, or should it be a NOT? (I'm just
asking to double-check, I haven't dug in and read enough of your test
to fully appreciate the chunk shown in the gitweb diff view.)

Thank *you* -- your persistence has paid off.

I think tomorrow it will be passing enough to merge over to 'master' ...
:-)

Great. Hopefully it will be useful to many.

I'm sure it will.
___
cmake-developers mailing list
cmake-developers@cmake.org
http://public.kitware.com/cgi-bin/mailman/listinfo/cmake-developers

Re: [cmake-developers] GenerateExportHeader macro in CMake?

2011-08-23 Thread David Cole

On Thu, Aug 11, 2011 at 11:30 AM, Stephen Kelly steve...@gmail.com wrote:
David Cole wrote:

Stephen,

Welcome to CMake development :-)

Thanks :),

Brad and I are both fairly busy this week. Please continue making fixes
and merging them into 'next' for dashboard testing as you are able. You're
getting closer.

It looks like you already made a similar change to what Bill was asking us
to push...?

Yep, I pushed a similar change to that, and it passed in the continuous
build already.

Let me know if you would like me to try anything on a specific machine
here at Kitware for you. I have access to all the Kitware dashboard
machines, and even some of the non-Kitware dashboard machines.

Thanks for your effort... We've all been there when adding something brand
new into CMake.

I think it will be mostly green again tomorrow in the nightlies.

I think the deprecated attributes are excluded a bit too pessimistically at
the moment, but I'll wait until it's all green before messing around with
that.

Keep up the good work!

Thanks :),

Steve.

David Cole
Kitware, Inc.

___
cmake-developers mailing list
cmake-developers@cmake.org
http://public.kitware.com/cgi-bin/mailman/listinfo/cmake-developers

Hi Stephen,

So... this topic is in 'next', and today, it only fails on 4
dashboards in total. I will try to get output for you on the wcl
(Watcom) compiler dashboard -- since it timed out, there is no test
output reported on CDash...

http://www.cdash.org/CDash/testSummary.php?project=1name=Module.GenerateExportHeaderdate=2011-08-23

Can you take a look at the CDash output of the other platforms where
this test is still failing, and recommend changes that might eliminate
all the test failures?

We can't really merge this to 'master' and include it in the upcoming
CMake 2.8.6 until we've resolved, worked-around or avoided the test
failures on the remaining dashboard submissions.

Thanks,
David Cole
Kitware, Inc.
___
cmake-developers mailing list
cmake-developers@cmake.org
http://public.kitware.com/cgi-bin/mailman/listinfo/cmake-developers

Re: [cmake-developers] GenerateExportHeader macro in CMake?

2011-08-23 Thread Stephen Kelly

David Cole wrote:
Hi Stephen,