Bug#758725: package mrpt_1:1.2.1-1 FTBFS on i386, amd64, mips*
Hi Jose Luis, I can confirm that the head of git (from Sep, 10, 2014) has been built successfully on mips and mipsel. All tests executed successfully. Please note that changes from Sep 11 and 12 have not been included in this and that I have set BUILD_ASSIMP=OFF for mips and mipsel. Could you please prepare new version of mrpt package for Debian? Thank you! Jurica From: Jose Luis Blanco [joseluisblan...@gmail.com] Sent: 09 September 2014 02:18 To: Jurica Stanojkovic Cc: 758...@bugs.debian.org Subject: Re: Bug#758725: package mrpt_1:1.2.1-1 FTBFS on i386, amd64, mips* Thanks for the investigation! Following that line, I think I have solved all de-serialization problems in GIT head: https://github.com/jlblancoc/mrpt/commits/master The two relevant patches are: 1) https://github.com/jlblancoc/mrpt/commit/7b1536c7f43ddc9cc949407f71fde0c8db0ecf5b 2) https://github.com/jlblancoc/mrpt/commit/d7fe726eb896b1a9d453e3b2d53252b25dbc3918 Let me know if this finally solves the issue. JL -- To UNSUBSCRIBE, email to debian-bugs-dist-requ...@lists.debian.org with a subject of unsubscribe. Trouble? Contact listmas...@lists.debian.org
Bug#758725: package mrpt_1:1.2.1-1 FTBFS on i386, amd64, mips*
Great news! Sure. I'm preparing it, after some checking will submit it. Thanks. JL On Fri, Sep 12, 2014 at 10:17 AM, Jurica Stanojkovic jurica.stanojko...@imgtec.com wrote: Hi Jose Luis, I can confirm that the head of git (from Sep, 10, 2014) has been built successfully on mips and mipsel. All tests executed successfully. Please note that changes from Sep 11 and 12 have not been included in this and that I have set BUILD_ASSIMP=OFF for mips and mipsel. Could you please prepare new version of mrpt package for Debian? Thank you! Jurica From: Jose Luis Blanco [joseluisblan...@gmail.com] Sent: 09 September 2014 02:18 To: Jurica Stanojkovic Cc: 758...@bugs.debian.org Subject: Re: Bug#758725: package mrpt_1:1.2.1-1 FTBFS on i386, amd64, mips* Thanks for the investigation! Following that line, I think I have solved all de-serialization problems in GIT head: https://github.com/jlblancoc/mrpt/commits/master The two relevant patches are: 1) https://github.com/jlblancoc/mrpt/commit/7b1536c7f43ddc9cc949407f71fde0c8db0ecf5b 2) https://github.com/jlblancoc/mrpt/commit/d7fe726eb896b1a9d453e3b2d53252b25dbc3918 Let me know if this finally solves the issue. JL -- ___ Dr. Jose-Luis Blanco-Claraco CITE-IV 1.05 Universidad de Almería, Departamento de Ingeniería 04120 Almería (Spain) http://www.ual.es/~jlblanco/ ___ -- To UNSUBSCRIBE, email to debian-bugs-dist-requ...@lists.debian.org with a subject of unsubscribe. Trouble? Contact listmas...@lists.debian.org
Bug#758725: package mrpt_1:1.2.1-1 FTBFS on i386, amd64, mips*
Hello Jose Luis, On big endian. while executing MonteCarlo2D.RunSampleDataset test first problem is found here: https://github.com/jlblancoc/mrpt/blob/master/libs/base/src/utils/CStream.cpp After the line 405 version_old have value that is in wrong endian format (little endian). Second problem is here: https://github.com/jlblancoc/mrpt/blob/master/libs/maps/src/maps/COccupancyGridMap2D_insert.cpp After the line 168 curRange have value that is in wrong endian format (little endian). For now I have used mrpt::utils::reverseBytesInPlace( ) to fix wrong endian values and I am trying to investigate problem further. Afrer resolving those problems MonteCarlo2D.RunSampleDataset test fail with message: *Warning: Test failed. Will give it another chance, since after all it's nondeterministic! Can you please look and comment at these issues? Thank you! Jurica From: Jurica Stanojkovic Sent: 27 August 2014 17:36 To: Jose Luis Blanco Cc: 758...@bugs.debian.org Subject: RE: Bug#758725: package mrpt_1:1.2.1-1 FTBFS on i386, amd64, mips* I have enabled and run test TEST(ReactiveNavTests, LoadNavLogFile) on mips big endian. Test does not fail on mips. Maybe, this test could be available for big endian arch as well. Regards, Jurica -- To UNSUBSCRIBE, email to debian-bugs-dist-requ...@lists.debian.org with a subject of unsubscribe. Trouble? Contact listmas...@lists.debian.org
Bug#758725: package mrpt_1:1.2.1-1 FTBFS on i386, amd64, mips*
Thanks for the investigation! Following that line, I think I have solved all de-serialization problems in GIT head: https://github.com/jlblancoc/mrpt/commits/master The two relevant patches are: 1) https://github.com/jlblancoc/mrpt/commit/7b1536c7f43ddc9cc949407f71fde0c8db0ecf5b 2) https://github.com/jlblancoc/mrpt/commit/d7fe726eb896b1a9d453e3b2d53252b25dbc3918 Let me know if this finally solves the issue. JL On Mon, Sep 8, 2014 at 5:26 PM, Jurica Stanojkovic jurica.stanojko...@imgtec.com wrote: Hello Jose Luis, On big endian. while executing MonteCarlo2D.RunSampleDataset test first problem is found here: https://github.com/jlblancoc/mrpt/blob/master/libs/base/src/utils/CStream.cpp After the line 405 version_old have value that is in wrong endian format (little endian). Second problem is here: https://github.com/jlblancoc/mrpt/blob/master/libs/maps/src/maps/COccupancyGridMap2D_insert.cpp After the line 168 curRange have value that is in wrong endian format (little endian). For now I have used mrpt::utils::reverseBytesInPlace( ) to fix wrong endian values and I am trying to investigate problem further. Afrer resolving those problems MonteCarlo2D.RunSampleDataset test fail with message: *Warning: Test failed. Will give it another chance, since after all it's nondeterministic! Can you please look and comment at these issues? Thank you! Jurica -- To UNSUBSCRIBE, email to debian-bugs-dist-requ...@lists.debian.org with a subject of unsubscribe. Trouble? Contact listmas...@lists.debian.org
Bug#758725: package mrpt_1:1.2.1-1 FTBFS on i386, amd64, mips*
Hello, I can confirm that the current head of git (taken on monday) has been built successfully on mips as well. All tests executed without an error. TEST(ReactiveNavTests, LoadNavLogFile) and TEST(MonteCarlo2D, RunSampleDataset) were skipped on big endian. Also TEST(Synch, CSemaphore_named ) has returned the mesage: *Skipping test* It seems the kernel doesn't support named semaphores in this platform. Regards, Jurica From: Jurica Stanojkovic Sent: 25 August 2014 15:52 To: Jose Luis Blanco Cc: 758...@bugs.debian.org Subject: RE: Bug#758725: package mrpt_1:1.2.1-1 FTBFS on i386, amd64, mips* Good! Just for curiosity: are all other unit tests passing? I have re-run tests again to be sure and they all pass. TEST(ReactiveNavTests, LoadNavLogFile) and TEST(MonteCarlo2D, RunSampleDataset) are skipped on big endian. Also TEST(Synch, CSemaphore_named ) has returned the mesage: *Skipping test* It seems the kernel doesn't support named semaphores in this platform. It would be great if you could confirm that the current head of git master is building correctly in mips... It has been built correctly on mipsel over weekend. On mips I am trying to build package. It will take more time on mips, because mips machine that I use is slower then the mipsel one. Regards, Jurica From: Jose Luis Blanco [joseluisblan...@gmail.com] Sent: 22 August 2014 17:58 To: Jurica Stanojkovic Cc: 758...@bugs.debian.org Subject: Re: Bug#758725: package mrpt_1:1.2.1-1 FTBFS on i386, amd64, mips* With these changes I was able to build package on mips* Good! Just for curiosity: are all other unit tests passing? Because that means that de-serialization is actually working (it's used in other tests, like in ReactiveNavTests.LoadNavLogFile). I have finally solved the pragma pack problem properly, with these 2 additional commits: 4) https://github.com/jlblancoc/mrpt/commit/cd4878874b930e151deb6d89577c215c687b80b3 5) https://github.com/jlblancoc/mrpt/commit/b52d65b5a0bcc552e7a7c6d420f30be623d8fd38 It would be great if you could confirm that the current head of git master is building correctly in mips... I guess it may take a long time to build in mips, so thank you very much for your tests!! Cheers, JL -- To UNSUBSCRIBE, email to debian-bugs-dist-requ...@lists.debian.org with a subject of unsubscribe. Trouble? Contact listmas...@lists.debian.org
Bug#758725: package mrpt_1:1.2.1-1 FTBFS on i386, amd64, mips*
Thank you very much!! On Wed, Aug 27, 2014 at 1:53 PM, Jurica Stanojkovic jurica.stanojko...@imgtec.com wrote: Hello, I can confirm that the current head of git (taken on monday) has been built successfully on mips as well. All tests executed without an error. TEST(ReactiveNavTests, LoadNavLogFile) and TEST(MonteCarlo2D, RunSampleDataset) were skipped on big endian. Also TEST(Synch, CSemaphore_named ) has returned the mesage: *Skipping test* It seems the kernel doesn't support named semaphores in this platform. Regards, Jurica -- To UNSUBSCRIBE, email to debian-bugs-dist-requ...@lists.debian.org with a subject of unsubscribe. Trouble? Contact listmas...@lists.debian.org
Bug#758725: package mrpt_1:1.2.1-1 FTBFS on i386, amd64, mips*
I have enabled and run test TEST(ReactiveNavTests, LoadNavLogFile) on mips big endian. Test does not fail on mips. Maybe, this test could be available for big endian arch as well. Regards, Jurica -- To UNSUBSCRIBE, email to debian-bugs-dist-requ...@lists.debian.org with a subject of unsubscribe. Trouble? Contact listmas...@lists.debian.org
Bug#758725: package mrpt_1:1.2.1-1 FTBFS on i386, amd64, mips*
Good! Just for curiosity: are all other unit tests passing? I have re-run tests again to be sure and they all pass. TEST(ReactiveNavTests, LoadNavLogFile) and TEST(MonteCarlo2D, RunSampleDataset) are skipped on big endian. Also TEST(Synch, CSemaphore_named ) has returned the mesage: *Skipping test* It seems the kernel doesn't support named semaphores in this platform. It would be great if you could confirm that the current head of git master is building correctly in mips... It has been built correctly on mipsel over weekend. On mips I am trying to build package. It will take more time on mips, because mips machine that I use is slower then the mipsel one. Regards, Jurica From: Jose Luis Blanco [joseluisblan...@gmail.com] Sent: 22 August 2014 17:58 To: Jurica Stanojkovic Cc: 758...@bugs.debian.org Subject: Re: Bug#758725: package mrpt_1:1.2.1-1 FTBFS on i386, amd64, mips* With these changes I was able to build package on mips* Good! Just for curiosity: are all other unit tests passing? Because that means that de-serialization is actually working (it's used in other tests, like in ReactiveNavTests.LoadNavLogFile). I have finally solved the pragma pack problem properly, with these 2 additional commits: 4) https://github.com/jlblancoc/mrpt/commit/cd4878874b930e151deb6d89577c215c687b80b3 5) https://github.com/jlblancoc/mrpt/commit/b52d65b5a0bcc552e7a7c6d420f30be623d8fd38 It would be great if you could confirm that the current head of git master is building correctly in mips... I guess it may take a long time to build in mips, so thank you very much for your tests!! Cheers, JL -- To UNSUBSCRIBE, email to debian-bugs-dist-requ...@lists.debian.org with a subject of unsubscribe. Trouble? Contact listmas...@lists.debian.org
Bug#758725: package mrpt_1:1.2.1-1 FTBFS on i386, amd64, mips*
I will try to fix de-serialization problems on big endian.
For now i have tried to build package with your patches
1)
https://github.com/jlblancoc/mrpt/commit/9a878c0950edd8ca7de6f4584f920c4e0db05ecc
2)
https://github.com/jlblancoc/mrpt/commit/289498fc34703c2fcf6b61e72c3311144e5fb570
and ifdef workaround
3)
https://github.com/jlblancoc/mrpt/commit/c06ed1c384ba2e719c52f65a1a49bddcf7d261e5
and the patch
4) mips-patch.diff (attached in this mail)
With these changes I was able to build package on mips*
For strict alignment, as far as I know, there is no macro that could be used
for detecting if a strict alignment is necessary.
I think that sparc also has alignment rules.
(maybe the best way to be sure about this is to ask porters for other
architectures to help with this issue).
Regards,
Jurica
From: Jose Luis Blanco [joseluisblan...@gmail.com]
Sent: 21 August 2014 19:11
To: Jurica Stanojkovic
Cc: 758...@bugs.debian.org
Subject: Re: Bug#758725: package mrpt_1:1.2.1-1 FTBFS on i386, amd64, mips*
So, the #ifdef workaround was incorrectly placed, to start with!
This patch will solve it:
https://github.com/jlblancoc/mrpt/commit/c06ed1c384ba2e719c52f65a1a49bddcf7d261e5
But obviously, that's not the ideal solution because that's resigning
trying to fix de-serialization problems!
I didn't know that issue about #pragma pack, it will be a problem.
Let me check it more in depth, but I'm sure the struct pack'ing is
taken for granted in some parts of the code. Actually, there's even
one unit test checking it...
Please, let me know whatever other unit tests don't pass in MIPS while
I think what to do with the pragmas.
BTW: Is there any macro to detect architectures with strict alignment rules?
JL
diff -upNr mrpt-1.2.1-old/libs/base/include/mrpt/math/lightweight_geom_data.h mrpt-1.2.1/libs/base/include/mrpt/math/lightweight_geom_data.h
--- mrpt-1.2.1-old/libs/base/include/mrpt/math/lightweight_geom_data.h 2014-07-10 16:18:40.0 +
+++ mrpt-1.2.1/libs/base/include/mrpt/math/lightweight_geom_data.h 2014-08-01 17:40:08.0 +
@@ -25,7 +25,9 @@ namespace math {
* @{ */
//Pragma defined to ensure no structure packing
+#ifndef __mips__
#pragma pack(push,1)
+#endif
//Set of typedefs for lightweight geometric items.
/**
* Lightweight 2D point. Allows coordinate access using [] operator.
@@ -238,7 +240,9 @@ namespace math {
static size_t size() { return 3; }
};
+#ifndef __mips__
#pragma pack(pop) // NOTE: Don't force TPoint3Df to be mem aligned (may break CPU mem access alignment in ARM)
+#endif
/** Lightweight 3D point (float version).
* \sa mrpt::poses::CPoint3D, mrpt::math::TPoint3D
@@ -261,7 +265,9 @@ namespace math {
};
+#ifndef __mips__
#pragma pack(push,1) //Pragma defined to ensure no structure packing
+#endif
/**
* Lightweight 3D point. Allows coordinate access using [] operator.
* \sa mrpt::poses::CPoint3D, mrpt::math::TPoint3Df
@@ -550,7 +556,9 @@ namespace math {
void fromString(const std::string s);
static size_t size() { return 7; }
};
+#ifndef __mips__
#pragma pack(pop)
+#endif
// Text streaming functions:
std::ostream BASE_IMPEXP operator (std::ostream o, const TPoint2D p);
@@ -621,7 +629,9 @@ namespace math {
struct BASE_IMPEXP TObject3D;
//Pragma defined to ensure no structure packing
+#ifndef __mips__
#pragma pack(push,1)
+#endif
/**
* 2D segment, consisting of two points.
* \sa TSegment3D,TLine2D,TPolygon2D,TPoint2D
@@ -761,7 +771,9 @@ namespace math {
bool operator(const TSegment3D s) const;
};
+#ifndef __mips__
#pragma pack(pop)
+#endif
inline bool operator==(const TSegment2D s1,const TSegment2D s2) {
return (s1.point1==s2.point1)(s1.point2==s2.point2);
Bug#758725: package mrpt_1:1.2.1-1 FTBFS on i386, amd64, mips*
With these changes I was able to build package on mips* Good! Just for curiosity: are all other unit tests passing? Because that means that de-serialization is actually working (it's used in other tests, like in ReactiveNavTests.LoadNavLogFile). I have finally solved the pragma pack problem properly, with these 2 additional commits: 4) https://github.com/jlblancoc/mrpt/commit/cd4878874b930e151deb6d89577c215c687b80b3 5) https://github.com/jlblancoc/mrpt/commit/b52d65b5a0bcc552e7a7c6d420f30be623d8fd38 It would be great if you could confirm that the current head of git master is building correctly in mips... I guess it may take a long time to build in mips, so thank you very much for your tests!! Cheers, JL -- To UNSUBSCRIBE, email to debian-bugs-dist-requ...@lists.debian.org with a subject of unsubscribe. Trouble? Contact listmas...@lists.debian.org
Bug#758725: package mrpt_1:1.2.1-1 FTBFS on i386, amd64, mips*
Hi Jose Luis, Thank you for the quick response! I can confirm that patches that you have proposed have resolved build issue on i386. I am working on resolving build issues for mips and mipsel. On mips and mipsel I have looked into https://github.com/jlblancoc/mrpt/blob/master/libs/slam/src/slam/CMonteCarloLocalization2D_unittest.cpp function void run_test_pf_localization(CPose2D meanPose, CMatrixDouble33 cov) on big endian is not doing anything, but run_test_pf_localization is used in TEST(MonteCarlo2D, RunSampleDataset). #if MRPT_IS_BIG_ENDIAN MRPT_TODO(Debug this issue in big endian platforms) return; // Skip this test for now #endif That is causing the following error: https://buildd.debian.org/status/fetch.php?pkg=mrptarch=mipsver=1%3A1.2.1-1stamp=1406312477 I will look further into function run_test_pf_localization. There is also .h file: https://github.com/jlblancoc/mrpt/blob/master/libs/base/include/mrpt/math/lightweight_geom_data.h containing #pragma pack(push,1) directive for structures that have float and double arguments. I think that this is causing unaligned access on architectures with strict alignment rules: https://buildd.debian.org/status/fetch.php?pkg=mrptarch=mipselver=1%3A1.2.1-1stamp=1406169915 For mips* architecture double should be 8 byte aligned and float should be 4 byte aligned. Can we disable #pragma pack(push,1) in this file for architectures with strict alignment rules? Any comments are welcome. Thanks! Jurica From: Jose Luis Blanco [joseluisblan...@gmail.com] Sent: 21 August 2014 01:44 To: Jurica Stanojkovic; 758...@bugs.debian.org Subject: Re: Bug#758725: package mrpt_1:1.2.1-1 FTBFS on i386, amd64, mips* Hi Jurica, Please, hopefully these two patches can solve all named semaphore errors: 1) https://github.com/jlblancoc/mrpt/commit/9a878c0950edd8ca7de6f4584f920c4e0db05ecc 2) https://github.com/jlblancoc/mrpt/commit/289498fc34703c2fcf6b61e72c3311144e5fb570 Cheers, JL On Wed, Aug 20, 2014 at 7:40 PM, Jose Luis Blanco joseluisblan...@gmail.com wrote: Thanks a lot for your interest in fixing this, Jurica! I'm the upstream maintainer of mrpt, and was becoming desperate trying to debug in mips* for my lack of experience in cross-buidling, qemu, etc. So, on the build errors in mips (and other big-endian platforms): I *suspect* that some errors may be caused by missing byte-reordering in serialization, via mrpt::utils::CStream and/or the affected classes methods named readFromStream() and writeToStream(), so that's where you could first look at while debugging in mips*. At least, that's what I suspect, but couldn't check it because can't debug in those platforms! :-( About the other errors in this present bug: Yes, I'll study them. In the meanwhile, you can avoid them by, instead of running make test, running: tests/test_mrpt_base --gtest_filter=-*CSemaphore* to run all tests in the lib mrpt-base, excepting those that give you problems. For the other test sets, you can also try, for example: make run_tests_mrpt_opengl and so on. Hit TAB after make run_tests_mrpt_ to see the full list. Hope it helps! JL On Wed, Aug 20, 2014 at 6:12 PM, Jurica Stanojkovic jurica.stanojko...@imgtec.com wrote: Package: mrpt Version: 1:1.2.1-1 Severity: serious Tags: sid Justification: FTBFS User: debian-mips-dev-disc...@lists.alioth.debian.org Hello, package mrpt_1:1.2.1-1 FTBFS on i386, amd64, mips* with the following message, but was previously built successfully on i386, amd64: i386 log: [--] 5 tests from Synch [ RUN ] Synch.CSemaphore_named_6t_1init Exception in [auxiliary_thread_launcher_LIN/WIN]!!!: === MRPT EXCEPTION = mrpt::synch::CSemaphore::CSemaphore(unsigned int, unsigned int, const string), line 81: Creating semaphore (name='mrpt-demo-sem1') raised error: Function not implemented mrpt::synch::CSemaphore::CSemaphore(unsigned int, unsigned int, const string), line 87: /«PKGBUILDDIR»/libs/base/src/synch/CCriticalSection_unittest.cpp:53: Failure Value of: false Actual: false Expected: true Thread didn't finished in timeout! While testing: CSemaphore named: #threads=6, init count=1 [ FAILED ] Synch.CSemaphore_named_6t_1init (5000 ms) [ RUN ] Synch.CSemaphore_named_10t_5init Exception in [auxiliary_thread_launcher_LIN/WIN]!!!: === MRPT EXCEPTION = mrpt::synch::CSemaphore::CSemaphore(unsigned int, unsigned int, const string), line 81: Creating semaphore (name='mrpt-demo-sem1') raised error: Function not implemented mrpt::synch::CSemaphore::CSemaphore(unsigned int, unsigned int, const string), line 87: /«PKGBUILDDIR»/libs/base/src/synch/CCriticalSection_unittest.cpp:53: Failure Value of: false Actual: false Expected: true Thread didn't finished in timeout! While testing: CSemaphore named: #threads=10, init count=5 [ FAILED ] Synch.CSemaphore_named_10t_5init (5000 ms
Bug#758725: package mrpt_1:1.2.1-1 FTBFS on i386, amd64, mips*
So, the #ifdef workaround was incorrectly placed, to start with! This patch will solve it: https://github.com/jlblancoc/mrpt/commit/c06ed1c384ba2e719c52f65a1a49bddcf7d261e5 But obviously, that's not the ideal solution because that's resigning trying to fix de-serialization problems! I didn't know that issue about #pragma pack, it will be a problem. Let me check it more in depth, but I'm sure the struct pack'ing is taken for granted in some parts of the code. Actually, there's even one unit test checking it... Please, let me know whatever other unit tests don't pass in MIPS while I think what to do with the pragmas. BTW: Is there any macro to detect architectures with strict alignment rules? JL On Thu, Aug 21, 2014 at 5:10 PM, Jurica Stanojkovic jurica.stanojko...@imgtec.com wrote: Hi Jose Luis, Thank you for the quick response! I can confirm that patches that you have proposed have resolved build issue on i386. I am working on resolving build issues for mips and mipsel. On mips and mipsel I have looked into https://github.com/jlblancoc/mrpt/blob/master/libs/slam/src/slam/CMonteCarloLocalization2D_unittest.cpp function void run_test_pf_localization(CPose2D meanPose, CMatrixDouble33 cov) on big endian is not doing anything, but run_test_pf_localization is used in TEST(MonteCarlo2D, RunSampleDataset). #if MRPT_IS_BIG_ENDIAN MRPT_TODO(Debug this issue in big endian platforms) return; // Skip this test for now #endif That is causing the following error: https://buildd.debian.org/status/fetch.php?pkg=mrptarch=mipsver=1%3A1.2.1-1stamp=1406312477 I will look further into function run_test_pf_localization. There is also .h file: https://github.com/jlblancoc/mrpt/blob/master/libs/base/include/mrpt/math/lightweight_geom_data.h containing #pragma pack(push,1) directive for structures that have float and double arguments. I think that this is causing unaligned access on architectures with strict alignment rules: https://buildd.debian.org/status/fetch.php?pkg=mrptarch=mipselver=1%3A1.2.1-1stamp=1406169915 For mips* architecture double should be 8 byte aligned and float should be 4 byte aligned. Can we disable #pragma pack(push,1) in this file for architectures with strict alignment rules? Any comments are welcome. Thanks! Jurica -- To UNSUBSCRIBE, email to debian-bugs-dist-requ...@lists.debian.org with a subject of unsubscribe. Trouble? Contact listmas...@lists.debian.org
Bug#758725: package mrpt_1:1.2.1-1 FTBFS on i386, amd64, mips*
Package: mrpt Version: 1:1.2.1-1 Severity: serious Tags: sid Justification: FTBFS User: debian-mips-dev-disc...@lists.alioth.debian.org Hello, package mrpt_1:1.2.1-1 FTBFS on i386, amd64, mips* with the following message, but was previously built successfully on i386, amd64: i386 log: [--] 5 tests from Synch [ RUN ] Synch.CSemaphore_named_6t_1init Exception in [auxiliary_thread_launcher_LIN/WIN]!!!: === MRPT EXCEPTION = mrpt::synch::CSemaphore::CSemaphore(unsigned int, unsigned int, const string), line 81: Creating semaphore (name='mrpt-demo-sem1') raised error: Function not implemented mrpt::synch::CSemaphore::CSemaphore(unsigned int, unsigned int, const string), line 87: /«PKGBUILDDIR»/libs/base/src/synch/CCriticalSection_unittest.cpp:53: Failure Value of: false Actual: false Expected: true Thread didn't finished in timeout! While testing: CSemaphore named: #threads=6, init count=1 [ FAILED ] Synch.CSemaphore_named_6t_1init (5000 ms) [ RUN ] Synch.CSemaphore_named_10t_5init Exception in [auxiliary_thread_launcher_LIN/WIN]!!!: === MRPT EXCEPTION = mrpt::synch::CSemaphore::CSemaphore(unsigned int, unsigned int, const string), line 81: Creating semaphore (name='mrpt-demo-sem1') raised error: Function not implemented mrpt::synch::CSemaphore::CSemaphore(unsigned int, unsigned int, const string), line 87: /«PKGBUILDDIR»/libs/base/src/synch/CCriticalSection_unittest.cpp:53: Failure Value of: false Actual: false Expected: true Thread didn't finished in timeout! While testing: CSemaphore named: #threads=10, init count=5 [ FAILED ] Synch.CSemaphore_named_10t_5init (5000 ms) [ RUN ] Synch.CriticalSections_Simple [ OK ] Synch.CriticalSections_Simple (1 ms) [ RUN ] Synch.CriticalSections_NoDoubleEnter [ OK ] Synch.CriticalSections_NoDoubleEnter (1 ms) [ RUN ] Synch.CriticalSections_Multi [ OK ] Synch.CriticalSections_Multi (1557 ms) [--] 5 tests from Synch (11560 ms total) ... [--] Global test environment tear-down [==] 138 tests from 23 test cases ran. (11643 ms total) [ PASSED ] 136 tests. [ FAILED ] 2 tests, listed below: [ FAILED ] Synch.CSemaphore_named_6t_1init [ FAILED ] Synch.CSemaphore_named_10t_5init 2 FAILED TESTS make[4]: *** [tests/CMakeFiles/run_tests_mrpt_base] Error 1 tests/CMakeFiles/run_tests_mrpt_base.dir/build.make:52: recipe for target 'tests/CMakeFiles/run_tests_mrpt_base' failed make[4]: Leaving directory '/«PKGBUILDDIR»/obj-i586-linux-gnu' make[3]: *** [tests/CMakeFiles/run_tests_mrpt_base.dir/all] Error 2 CMakeFiles/Makefile2:5096: recipe for target 'tests/CMakeFiles/run_tests_mrpt_base.dir/all' failed make[3]: Leaving directory '/«PKGBUILDDIR»/obj-i586-linux-gnu' make[2]: *** [tests/CMakeFiles/test.dir/rule] Error 2 CMakeFiles/Makefile2:5455: recipe for target 'tests/CMakeFiles/test.dir/rule' failed make[2]: Leaving directory '/«PKGBUILDDIR»/obj-i586-linux-gnu' make[1]: *** [test] Error 2 Makefile:1665: recipe for target 'test' failed make[1]: Leaving directory '/«PKGBUILDDIR»/obj-i586-linux-gnu' dh_auto_test: make -j1 test ARGS+=-j1 returned exit code 2 make: *** [build-arch] Error 2 debian/rules:44: recipe for target 'build-arch' failed dpkg-buildpackage: error: debian/rules build-arch gave error exit status 2 Could someone please help with this regression? I am working on resolving other build issues for mrpt package build on mips and mipsel, but I can not resolve them until this regression gets resolved. Thanks! Jurica
Bug#758725: package mrpt_1:1.2.1-1 FTBFS on i386, amd64, mips*
Thanks a lot for your interest in fixing this, Jurica! I'm the upstream maintainer of mrpt, and was becoming desperate trying to debug in mips* for my lack of experience in cross-buidling, qemu, etc. So, on the build errors in mips (and other big-endian platforms): I *suspect* that some errors may be caused by missing byte-reordering in serialization, via mrpt::utils::CStream and/or the affected classes methods named readFromStream() and writeToStream(), so that's where you could first look at while debugging in mips*. At least, that's what I suspect, but couldn't check it because can't debug in those platforms! :-( About the other errors in this present bug: Yes, I'll study them. In the meanwhile, you can avoid them by, instead of running make test, running: tests/test_mrpt_base --gtest_filter=-*CSemaphore* to run all tests in the lib mrpt-base, excepting those that give you problems. For the other test sets, you can also try, for example: make run_tests_mrpt_opengl and so on. Hit TAB after make run_tests_mrpt_ to see the full list. Hope it helps! JL On Wed, Aug 20, 2014 at 6:12 PM, Jurica Stanojkovic jurica.stanojko...@imgtec.com wrote: Package: mrpt Version: 1:1.2.1-1 Severity: serious Tags: sid Justification: FTBFS User: debian-mips-dev-disc...@lists.alioth.debian.org Hello, package mrpt_1:1.2.1-1 FTBFS on i386, amd64, mips* with the following message, but was previously built successfully on i386, amd64: i386 log: [--] 5 tests from Synch [ RUN ] Synch.CSemaphore_named_6t_1init Exception in [auxiliary_thread_launcher_LIN/WIN]!!!: === MRPT EXCEPTION = mrpt::synch::CSemaphore::CSemaphore(unsigned int, unsigned int, const string), line 81: Creating semaphore (name='mrpt-demo-sem1') raised error: Function not implemented mrpt::synch::CSemaphore::CSemaphore(unsigned int, unsigned int, const string), line 87: /«PKGBUILDDIR»/libs/base/src/synch/CCriticalSection_unittest.cpp:53: Failure Value of: false Actual: false Expected: true Thread didn't finished in timeout! While testing: CSemaphore named: #threads=6, init count=1 [ FAILED ] Synch.CSemaphore_named_6t_1init (5000 ms) [ RUN ] Synch.CSemaphore_named_10t_5init Exception in [auxiliary_thread_launcher_LIN/WIN]!!!: === MRPT EXCEPTION = mrpt::synch::CSemaphore::CSemaphore(unsigned int, unsigned int, const string), line 81: Creating semaphore (name='mrpt-demo-sem1') raised error: Function not implemented mrpt::synch::CSemaphore::CSemaphore(unsigned int, unsigned int, const string), line 87: /«PKGBUILDDIR»/libs/base/src/synch/CCriticalSection_unittest.cpp:53: Failure Value of: false Actual: false Expected: true Thread didn't finished in timeout! While testing: CSemaphore named: #threads=10, init count=5 [ FAILED ] Synch.CSemaphore_named_10t_5init (5000 ms) [ RUN ] Synch.CriticalSections_Simple [ OK ] Synch.CriticalSections_Simple (1 ms) [ RUN ] Synch.CriticalSections_NoDoubleEnter [ OK ] Synch.CriticalSections_NoDoubleEnter (1 ms) [ RUN ] Synch.CriticalSections_Multi [ OK ] Synch.CriticalSections_Multi (1557 ms) [--] 5 tests from Synch (11560 ms total) ... [--] Global test environment tear-down [==] 138 tests from 23 test cases ran. (11643 ms total) [ PASSED ] 136 tests. [ FAILED ] 2 tests, listed below: [ FAILED ] Synch.CSemaphore_named_6t_1init [ FAILED ] Synch.CSemaphore_named_10t_5init 2 FAILED TESTS make[4]: *** [tests/CMakeFiles/run_tests_mrpt_base] Error 1 tests/CMakeFiles/run_tests_mrpt_base.dir/build.make:52: recipe for target 'tests/CMakeFiles/run_tests_mrpt_base' failed make[4]: Leaving directory '/«PKGBUILDDIR»/obj-i586-linux-gnu' make[3]: *** [tests/CMakeFiles/run_tests_mrpt_base.dir/all] Error 2 CMakeFiles/Makefile2:5096: recipe for target 'tests/CMakeFiles/run_tests_mrpt_base.dir/all' failed make[3]: Leaving directory '/«PKGBUILDDIR»/obj-i586-linux-gnu' make[2]: *** [tests/CMakeFiles/test.dir/rule] Error 2 CMakeFiles/Makefile2:5455: recipe for target 'tests/CMakeFiles/test.dir/rule' failed make[2]: Leaving directory '/«PKGBUILDDIR»/obj-i586-linux-gnu' make[1]: *** [test] Error 2 Makefile:1665: recipe for target 'test' failed make[1]: Leaving directory '/«PKGBUILDDIR»/obj-i586-linux-gnu' dh_auto_test: make -j1 test ARGS+=-j1 returned exit code 2 make: *** [build-arch] Error 2 debian/rules:44: recipe for target 'build-arch' failed dpkg-buildpackage: error: debian/rules build-arch gave error exit status 2 Could someone please help with this regression? I am working on resolving other build issues for mrpt package build on mips and mipsel, but I can not resolve them until this regression gets resolved. Thanks! Jurica -- To UNSUBSCRIBE, email to debian-bugs-dist-requ...@lists.debian.org with a subject of unsubscribe. Trouble? Contact listmas...@lists.debian.org
Bug#758725: package mrpt_1:1.2.1-1 FTBFS on i386, amd64, mips*
Hi Jurica, Please, hopefully these two patches can solve all named semaphore errors: 1) https://github.com/jlblancoc/mrpt/commit/9a878c0950edd8ca7de6f4584f920c4e0db05ecc 2) https://github.com/jlblancoc/mrpt/commit/289498fc34703c2fcf6b61e72c3311144e5fb570 Cheers, JL On Wed, Aug 20, 2014 at 7:40 PM, Jose Luis Blanco joseluisblan...@gmail.com wrote: Thanks a lot for your interest in fixing this, Jurica! I'm the upstream maintainer of mrpt, and was becoming desperate trying to debug in mips* for my lack of experience in cross-buidling, qemu, etc. So, on the build errors in mips (and other big-endian platforms): I *suspect* that some errors may be caused by missing byte-reordering in serialization, via mrpt::utils::CStream and/or the affected classes methods named readFromStream() and writeToStream(), so that's where you could first look at while debugging in mips*. At least, that's what I suspect, but couldn't check it because can't debug in those platforms! :-( About the other errors in this present bug: Yes, I'll study them. In the meanwhile, you can avoid them by, instead of running make test, running: tests/test_mrpt_base --gtest_filter=-*CSemaphore* to run all tests in the lib mrpt-base, excepting those that give you problems. For the other test sets, you can also try, for example: make run_tests_mrpt_opengl and so on. Hit TAB after make run_tests_mrpt_ to see the full list. Hope it helps! JL On Wed, Aug 20, 2014 at 6:12 PM, Jurica Stanojkovic jurica.stanojko...@imgtec.com wrote: Package: mrpt Version: 1:1.2.1-1 Severity: serious Tags: sid Justification: FTBFS User: debian-mips-dev-disc...@lists.alioth.debian.org Hello, package mrpt_1:1.2.1-1 FTBFS on i386, amd64, mips* with the following message, but was previously built successfully on i386, amd64: i386 log: [--] 5 tests from Synch [ RUN ] Synch.CSemaphore_named_6t_1init Exception in [auxiliary_thread_launcher_LIN/WIN]!!!: === MRPT EXCEPTION = mrpt::synch::CSemaphore::CSemaphore(unsigned int, unsigned int, const string), line 81: Creating semaphore (name='mrpt-demo-sem1') raised error: Function not implemented mrpt::synch::CSemaphore::CSemaphore(unsigned int, unsigned int, const string), line 87: /«PKGBUILDDIR»/libs/base/src/synch/CCriticalSection_unittest.cpp:53: Failure Value of: false Actual: false Expected: true Thread didn't finished in timeout! While testing: CSemaphore named: #threads=6, init count=1 [ FAILED ] Synch.CSemaphore_named_6t_1init (5000 ms) [ RUN ] Synch.CSemaphore_named_10t_5init Exception in [auxiliary_thread_launcher_LIN/WIN]!!!: === MRPT EXCEPTION = mrpt::synch::CSemaphore::CSemaphore(unsigned int, unsigned int, const string), line 81: Creating semaphore (name='mrpt-demo-sem1') raised error: Function not implemented mrpt::synch::CSemaphore::CSemaphore(unsigned int, unsigned int, const string), line 87: /«PKGBUILDDIR»/libs/base/src/synch/CCriticalSection_unittest.cpp:53: Failure Value of: false Actual: false Expected: true Thread didn't finished in timeout! While testing: CSemaphore named: #threads=10, init count=5 [ FAILED ] Synch.CSemaphore_named_10t_5init (5000 ms) [ RUN ] Synch.CriticalSections_Simple [ OK ] Synch.CriticalSections_Simple (1 ms) [ RUN ] Synch.CriticalSections_NoDoubleEnter [ OK ] Synch.CriticalSections_NoDoubleEnter (1 ms) [ RUN ] Synch.CriticalSections_Multi [ OK ] Synch.CriticalSections_Multi (1557 ms) [--] 5 tests from Synch (11560 ms total) ... [--] Global test environment tear-down [==] 138 tests from 23 test cases ran. (11643 ms total) [ PASSED ] 136 tests. [ FAILED ] 2 tests, listed below: [ FAILED ] Synch.CSemaphore_named_6t_1init [ FAILED ] Synch.CSemaphore_named_10t_5init 2 FAILED TESTS make[4]: *** [tests/CMakeFiles/run_tests_mrpt_base] Error 1 tests/CMakeFiles/run_tests_mrpt_base.dir/build.make:52: recipe for target 'tests/CMakeFiles/run_tests_mrpt_base' failed make[4]: Leaving directory '/«PKGBUILDDIR»/obj-i586-linux-gnu' make[3]: *** [tests/CMakeFiles/run_tests_mrpt_base.dir/all] Error 2 CMakeFiles/Makefile2:5096: recipe for target 'tests/CMakeFiles/run_tests_mrpt_base.dir/all' failed make[3]: Leaving directory '/«PKGBUILDDIR»/obj-i586-linux-gnu' make[2]: *** [tests/CMakeFiles/test.dir/rule] Error 2 CMakeFiles/Makefile2:5455: recipe for target 'tests/CMakeFiles/test.dir/rule' failed make[2]: Leaving directory '/«PKGBUILDDIR»/obj-i586-linux-gnu' make[1]: *** [test] Error 2 Makefile:1665: recipe for target 'test' failed make[1]: Leaving directory '/«PKGBUILDDIR»/obj-i586-linux-gnu' dh_auto_test: make -j1 test ARGS+=-j1 returned exit code 2 make: *** [build-arch] Error 2 debian/rules:44: recipe for target 'build-arch' failed dpkg-buildpackage: error: debian/rules build-arch gave error exit status 2 Could someone please help with
