Bug#825845: mrpt: FTBFS on big-endian systems, with test suite errors

[Gianfranco Costamagna]

done thanks! :)

G.





Il Domenica 10 Luglio 2016 0:41, JOSE LUIS BLANCO CLARACO  ha 
scritto:
Hi again Gianfranco,

I just noticed a missing open bug regarding a FTBFS on sparc64. OK,
it's a weird platform... but I already had the fix upstream, it was
overlooked in the last set of patches.

I added a new patch for it in a new version 1.4.0-3 and just uploaded
it to Mentors [1]. It would be great if you could sponsor it, so the
package becomes bug-free...

Cheers and thanks again for everything!
JL


[1] https://mentors.debian.net/debian/pool/main/m/mrpt/mrpt_1.4.0-3.dsc

Bug#825845: mrpt: FTBFS on big-endian systems, with test suite errors

[JOSE LUIS BLANCO CLARACO]

Hi again Gianfranco,

I just noticed a missing open bug regarding a FTBFS on sparc64. OK,
it's a weird platform... but I already had the fix upstream, it was
overlooked in the last set of patches.

I added a new patch for it in a new version 1.4.0-3 and just uploaded
it to Mentors [1]. It would be great if you could sponsor it, so the
package becomes bug-free...

Cheers and thanks again for everything!
JL

[1] https://mentors.debian.net/debian/pool/main/m/mrpt/mrpt_1.4.0-3.dsc

Bug#825845: mrpt: FTBFS on big-endian systems, with test suite errors

[Gianfranco Costamagna]

Hi,
>All green! :-) See [1].


yep, I look to see it migrate in Ubuntu too, probably tomorrow
>Thank you so much for the push.


thanks to you for the fix and patch
>I guess that the second half of archs in [1] are not officially
>supported and it's not a big deal to have some failures on them,
>right?


you can put them in the long todolist :)
I don't care too much, but they are nice to fix, just don't loose
too much time if you see the failure is arch:specific and not something
just hidden somewhere else (sometimes you can spot general failures in 
particular
architectures)

thanks

G.

Bug#825845: mrpt: FTBFS on big-endian systems, with test suite errors

[JOSE LUIS BLANCO CLARACO]

All green! :-) See [1].

Thank you so much for the push.

I guess that the second half of archs in [1] are not officially
supported and it's not a big deal to have some failures on them,
right?

Best, have a nice weekend.
JL

[1] https://buildd.debian.org/status/package.php?p=mrpt

Bug#825845: mrpt: FTBFS on big-endian systems, with test suite errors

[JOSE LUIS BLANCO CLARACO]

Thanks so much!

Sure I will, every day learning something new...

Bug#825845: mrpt: FTBFS on big-endian systems, with test suite errors

[Gianfranco Costamagna]

Hi, I sponsored it, even if for a next time please do two things:
split the patch into two different files
one for armhf and one for BE failures.

"add-patch" is a good command to add single commits
wget github.com/project/commit.patch
(add the .patch)
and then add-patch of the file

but since they are upstream cherry-picks I don't care, they will disappear 
anyway.

More important was the missing change in control file
+Replaces: mrpt-apps (<< 1:1.3.2-1)
+Breaks: mrpt-apps (<< 1:1.3.2-1)

I added it and sponsored.

thanks! (fingers crossed)

G.



Il Sabato 9 Luglio 2016 0:31, JOSE LUIS BLANCO CLARACO  ha 
scritto:



> ok, rebased with current debian/unstable package and build good
>
> I did grab the package from unstable, added the commit above, and did a 
> complete build.
> It didn't fail on s390x, so I don't know how to trigger that failure.

Well, that's good news, I guess! Thank you for your time.

I have just cherry-picked some commits and applied them to 1.4.0-1
(current "unstable") to create 1.4.0-2 which should fix all FTBFS
bugs, hopefully...
The DSC is in [1]. This is my first debian package with patches (via
dquilt) but I think it's fine.

Perhaps you could consider sponsoring its upload to Debian so we can
close a few bugs and fix the package for big-endian platforms?

Best,
JL


[1] https://mentors.debian.net/debian/pool/main/m/mrpt/mrpt_1.4.0-2.dsc

Bug#825845: mrpt: FTBFS on big-endian systems, with test suite errors

[JOSE LUIS BLANCO CLARACO]

> ok, rebased with current debian/unstable package and build good
>
> I did grab the package from unstable, added the commit above, and did a 
> complete build.
> It didn't fail on s390x, so I don't know how to trigger that failure.

Well, that's good news, I guess! Thank you for your time.

I have just cherry-picked some commits and applied them to 1.4.0-1
(current "unstable") to create 1.4.0-2 which should fix all FTBFS
bugs, hopefully...
The DSC is in [1]. This is my first debian package with patches (via
dquilt) but I think it's fine.

Perhaps you could consider sponsoring its upload to Debian so we can
close a few bugs and fix the package for big-endian platforms?

Best,
JL

[1] https://mentors.debian.net/debian/pool/main/m/mrpt/mrpt_1.4.0-2.dsc

Bug#825845: mrpt: FTBFS on big-endian systems, with test suite errors

[Gianfranco Costamagna]

Hi,

>I think I've (properly) fixed the issues in big-endian architectures

>for one set of tests.
>It would be great if you could launch a build in a test machine to confirm 
>it...
>
>The patch is in [1]. You could test it over release 1.4.0 (*without*
>the latest patch which, if you recall it, just put a #if 0 around the
>failing tests!), or just grab the whole thing from git master [2].

ok, rebased with current debian/unstable package and build good

>From Debian logs, I see that there is one more test that fails in
>*some* big endian architectures. I'm almost sure it could be debugged
>by running it with gdb.
>In these last weeks I managed to create a system to run unit tests
>under gdb as part of the Debian build, but it's disabled by default
>because it caused problems in armhf.
>You could also uncomment it (see lines [3] of debian/rules) to see if
>we can get more useful info about potential failures...
>Just replace
>
>MRPT_TEST_TARGET = test
>
>with:
>
>MRPT_TEST_TARGET = test_gdb

ok



I did grab the package from unstable, added the commit above, and did a 
complete build.
It didn't fail on s390x, so I don't know how to trigger that failure.


thanks,
G.

On Mon, Jul 4, 2016 at 12:13 PM, Gianfranco Costamagna
 wrote:
> Hi,
>
>
>>lease, find the workaround (not solution!) commit in [1]. Please, if
>
>>possible, apply it directly over the current v1.4.0 Debian package to
>>unblock building in big endian platforms. It would be great if you
>>could sponsor the update in Debian, not only in Ubuntu.
>>
>>If I find spare time to work in a real solution, I'll contact you just
>>in case you could help me testing the patches in porter machines...
>
>
> I can sponsor whatever you give me, a dsc, a tarball of debian packaging 
> directory,
> whatever (a git snapshot)
>
>
> Right now, I applied the two commits as patches, and the fix for 
> breaks+replaces
> fields, and I uploaded it in Ubuntu (to check if everything is correct)
>
> I called it ~build2 [1], so on the Debian upload it will be overridden 
> automatically
> by the auto import robot
>
>
> here [2]
>
> [1] https://launchpad.net/ubuntu/+source/mrpt/1:1.4.0-1build2
>
> [2] 
> http://launchpadlibrarian.net/270793330/mrpt_1%3A1.4.0-1build1_1%3A1.4.0-1build2.diff.gz
>
> I'm looking the build logs, if you can give me a dsc file I'll sponsor it in 
> a matter of minutes.
>
> If you don't change the version, just send me a tarball of the debian 
> directory, it should be enough for me!
>
> thanks for "fixing" :)
>
> Gianfranco



-- 
___

Jose Luis Blanco-Claraco
CITE-IV 1.05
Universidad de Almería, Departamento de Ingeniería
04120 Almería (Spain)
http://www.ual.es/~jlblanco/

___

Bug#825845: mrpt: FTBFS on big-endian systems, with test suite errors

[JOSE LUIS BLANCO CLARACO]

Hi again Gianfranco,

I think I've (properly) fixed the issues in big-endian architectures
for one set of tests.
It would be great if you could launch a build in a test machine to confirm it...

The patch is in [1]. You could test it over release 1.4.0 (*without*
the latest patch which, if you recall it, just put a #if 0 around the
failing tests!), or just grab the whole thing from git master [2].

>From Debian logs, I see that there is one more test that fails in
*some* big endian architectures. I'm almost sure it could be debugged
by running it with gdb.
In these last weeks I managed to create a system to run unit tests
under gdb as part of the Debian build, but it's disabled by default
because it caused problems in armhf.
You could also uncomment it (see lines [3] of debian/rules) to see if
we can get more useful info about potential failures...
Just replace

MRPT_TEST_TARGET = test

with:

MRPT_TEST_TARGET = test_gdb


Thanks for your support!!

[1] https://github.com/MRPT/mrpt/commit/e791599a30a0b60b551ee3e31225609b1a798a39
[2] https://github.com/MRPT/mrpt
[3] 
https://github.com/MRPT/mrpt/blob/b346bbbadbcf0f582c078c9e975f0ad4e0125565/packaging/debian/rules#L23

On Mon, Jul 4, 2016 at 12:13 PM, Gianfranco Costamagna
 wrote:
> Hi,
>
>
>>lease, find the workaround (not solution!) commit in [1]. Please, if
>
>>possible, apply it directly over the current v1.4.0 Debian package to
>>unblock building in big endian platforms. It would be great if you
>>could sponsor the update in Debian, not only in Ubuntu.
>>
>>If I find spare time to work in a real solution, I'll contact you just
>>in case you could help me testing the patches in porter machines...
>
>
> I can sponsor whatever you give me, a dsc, a tarball of debian packaging 
> directory,
> whatever (a git snapshot)
>
>
> Right now, I applied the two commits as patches, and the fix for 
> breaks+replaces
> fields, and I uploaded it in Ubuntu (to check if everything is correct)
>
> I called it ~build2 [1], so on the Debian upload it will be overridden 
> automatically
> by the auto import robot
>
>
> here [2]
>
> [1] https://launchpad.net/ubuntu/+source/mrpt/1:1.4.0-1build2
>
> [2] 
> http://launchpadlibrarian.net/270793330/mrpt_1%3A1.4.0-1build1_1%3A1.4.0-1build2.diff.gz
>
> I'm looking the build logs, if you can give me a dsc file I'll sponsor it in 
> a matter of minutes.
>
> If you don't change the version, just send me a tarball of the debian 
> directory, it should be enough for me!
>
> thanks for "fixing" :)
>
> Gianfranco



-- 
___

Jose Luis Blanco-Claraco
CITE-IV 1.05
Universidad de Almería, Departamento de Ingeniería
04120 Almería (Spain)
http://www.ual.es/~jlblanco/
___

Bug#825845: mrpt: FTBFS on big-endian systems, with test suite errors

[Gianfranco Costamagna]

Hi,


>lease, find the workaround (not solution!) commit in [1]. Please, if

>possible, apply it directly over the current v1.4.0 Debian package to
>unblock building in big endian platforms. It would be great if you
>could sponsor the update in Debian, not only in Ubuntu.
>
>If I find spare time to work in a real solution, I'll contact you just
>in case you could help me testing the patches in porter machines...


I can sponsor whatever you give me, a dsc, a tarball of debian packaging 
directory,
whatever (a git snapshot)


Right now, I applied the two commits as patches, and the fix for breaks+replaces
fields, and I uploaded it in Ubuntu (to check if everything is correct)

I called it ~build2 [1], so on the Debian upload it will be overridden 
automatically
by the auto import robot


here [2]

[1] https://launchpad.net/ubuntu/+source/mrpt/1:1.4.0-1build2

[2] 
http://launchpadlibrarian.net/270793330/mrpt_1%3A1.4.0-1build1_1%3A1.4.0-1build2.diff.gz

I'm looking the build logs, if you can give me a dsc file I'll sponsor it in a 
matter of minutes.

If you don't change the version, just send me a tarball of the debian 
directory, it should be enough for me!

thanks for "fixing" :)

Gianfranco

Bug#825845: mrpt: FTBFS on big-endian systems, with test suite errors

[JOSE LUIS BLANCO CLARACO]

Hi,

Please, find the workaround (not solution!) commit in [1]. Please, if
possible, apply it directly over the current v1.4.0 Debian package to
unblock building in big endian platforms. It would be great if you
could sponsor the update in Debian, not only in Ubuntu.

If I find spare time to work in a real solution, I'll contact you just
in case you could help me testing the patches in porter machines...

Thanks for the support!

[1] https://github.com/MRPT/mrpt/commit/b247f14ffdaf8f31ec8cafefcc2a12ac1d210618

Bug#825845: mrpt: FTBFS on big-endian systems, with test suite errors

[Gianfranco Costamagna]

Hi,

>Sorry for the delay, but it's difficult for me to debug those tests
>because I can't run the tests in any local / remote machine...
>
>A few days after this bug report, I applied to become a DM (via my
>sponsor) in part as a way to be able to run these tests in Debian
>infraestructure.
>Do you know of another way to quickly do tests on those big-endian platforms?


I have an ongoing build of the current git master on zelenka.debian.org.
If you have tests/patches, ask me and I'll perform them
(build at 40%)

this is an s390x machine

>As an alternative, I can submit a patch to just skip those tests in
>big-endian platforms and leave this for future fix and in the
>meanwhile unblock the transition in Ubuntu.


if you can unblock the transition this would be great, just ping and I'll 
sponsor
it.
(I can also unblock just Ubuntu, and see Debian fixed properly, there is no 
real rush,
as soon as the transition lands in -release)

thanks!

G.


On Fri, Jul 1, 2016 at 12:07 PM, Gianfranco Costamagna
 wrote:
> Hi Jose, do you have any ETA for this issue?
>
> this is preventing the opencv decruft, and the transition in Ubuntu.
>
> thanks
>
> Gianfranco
>

Bug#825845: mrpt: FTBFS on big-endian systems, with test suite errors

[JOSE LUIS BLANCO CLARACO]

Hi Gianfranco ,

Sorry for the delay, but it's difficult for me to debug those tests
because I can't run the tests in any local / remote machine...

A few days after this bug report, I applied to become a DM (via my
sponsor) in part as a way to be able to run these tests in Debian
infraestructure.
Do you know of another way to quickly do tests on those big-endian platforms?

As an alternative, I can submit a patch to just skip those tests in
big-endian platforms and leave this for future fix and in the
meanwhile unblock the transition in Ubuntu.

JL


On Fri, Jul 1, 2016 at 12:07 PM, Gianfranco Costamagna
 wrote:
> Hi Jose, do you have any ETA for this issue?
>
> this is preventing the opencv decruft, and the transition in Ubuntu.
>
> thanks
>
> Gianfranco
>

Bug#825845: mrpt: FTBFS on big-endian systems, with test suite errors

[Gianfranco Costamagna]

Hi Jose, do you have any ETA for this issue?

this is preventing the opencv decruft, and the transition in Ubuntu.

thanks

Gianfranco



signature.asc
Description: OpenPGP digital signature

Bug#825845: mrpt: FTBFS on big-endian systems, with test suite errors

[Aaron M. Ucko]

Source: mrpt
Version: 1:1.4.0-1
Severity: serious
Justification: fails to build from source (but built successfully in the past)

Builds of mrpt on big-endian systems that have gotten as far as
running the test suite have been failing with test suite errors.
Specifically, on powerpc and s390x (both release architectures), the
sample_vlp16_dataset and sample_hdl32_dataset tests both failed:

  [ RUN  ] CVelodyneScanner.sample_vlp16_dataset
  [CObservationVelodyneScan] skipping invalid packet: block 0 header value is 
65518...[CObservationVelodyneScan] skipping invalid packet: block 11 header 
value is 
65518/«PKGBUILDDIR»/libs/hwdrivers/src/CVelodyneScanner_unittest.cpp:60: Failure
  Value of: 4
  Expected: nScans
  Which is: 250
  [  FAILED  ] CVelodyneScanner.sample_vlp16_dataset (25378 ms)
  [ RUN  ] CVelodyneScanner.sample_hdl32_dataset
  /«PKGBUILDDIR»/libs/hwdrivers/src/CVelodyneScanner_unittest.cpp:94: Failure
  Value of: 3
  Expected: nScans
  Which is: 349
  [  FAILED  ] CVelodyneScanner.sample_hdl32_dataset (35189 ms)

On hppa (admittedly not a release architecture), those two tests
failed as above, and so did parse_NMEA_RMC:

  [ RUN  ] CGPSInterface.parse_NMEA_RMC
  unknown file: Failure
  C++ exception with description "
  
   === MRPT EXCEPTION =
  void mrpt::system::timestampToParts(mrpt::system::TTimeStamp, 
mrpt::system::TTimeParts&, bool), line 120:
  Assert condition failed: sec_frac<1.0
  " thrown in the test body.
  [  FAILED  ] CGPSInterface.parse_NMEA_RMC (36 ms)

Could you please take a look?

Thanks!