Building gnuradio - those annoying swig doc errors

Hey list,

Ever build gnuradio with a lot of -j* and catch some error like “swig
doc generation failed, retrying…”? I believe this is a dependency
specification issue with the build system. Well, I think that I have
fixed it…

For those interested, give my josh_build_work branch a try
branch is josh_build_work on git://gnuradio.org/jblum.git
http://gnuradio.org/cgit/jblum.git/log/?h=josh_build_work

Let me know if the problem goes away for you too.
Feedback on this will be very helpful. Thanks!

-josh

On Tue, Jul 3, 2012 at 2:11 PM, Josh B. [email protected] wrote:

Feedback on this will be very helpful. Thanks!

Awesome stuff.

I cherry-picked the relevant commits onto a test branch based off maint
instead of master. There were a couple minor conflicts easily fixed.
So
far -j8 works fine, and I’m testing -j now.

This has been pushed to branch ‘test-doc-fix-maint’ on the main GNU
Radio
repository, for easier testing.

Johnathan

On Tue, Jul 3, 2012 at 2:35 PM, Johnathan C. <
[email protected]> wrote:

I cherry-picked the relevant commits onto a test branch based off maint
instead of master. There were a couple minor conflicts easily fixed. So
far -j8 works fine, and I’m testing -j now.

There still seems to be an issue triggered when using ‘make -j’. The end
result is that the SWIG generated gnuradio_core*.so files have no
actual
code content, just overhead, and end up only a few kilobytes in length.
Then running make again causes the SWIG generated cxx files to get
compiled, but the output .so files are still broken (all the QA fails).

Notably, none of the original SWIG doc generation stuff is failing
anymore,
so that appears to be fixed, but clearly there is more going on. I’ve
seen
the above issue in the past with -j8 and -j16 on rare occasions, so this
issue has been around awhile.

Johnathan

On Tue, Jul 3, 2012 at 3:31 PM, Josh B. [email protected] wrote:

So is this new, ie added by these swig fixes? I have never seen this one
before.

I don’t think the .so issue is caused by the swig fixes. Doing ‘make
-j’
on master fails the same way.

But the swig doc issues that you were trying to fix do indeed appear to
have been fixed.

Johnathan

On 07/03/2012 03:14 PM, Johnathan C. wrote:

result is that the SWIG generated gnuradio_core*.so files have no actual
code content, just overhead, and end up only a few kilobytes in length.
Then running make again causes the SWIG generated cxx files to get
compiled, but the output .so files are still broken (all the QA fails).

So is this new, ie added by these swig fixes? I have never seen this one
before.

Notably, none of the original SWIG doc generation stuff is failing anymore,
so that appears to be fixed, but clearly there is more going on. I’ve seen
the above issue in the past with -j8 and -j16 on rare occasions, so this
issue has been around awhile.

Just to confirm. It is not failing anymore. We have forward progress
(or a little backward progress as well)?

-josh

On Tue, Jul 3, 2012 at 9:34 PM, Johnathan C.
[email protected] wrote:

Both the above issues have been fixed and pushed to maint->master->next.

Johnathan

Thanks for fixing this, guys!

Tom

On Tue, Jul 3, 2012 at 3:42 PM, Johnathan C. <
[email protected]> wrote:

On Tue, Jul 3, 2012 at 3:31 PM, Josh B. [email protected] wrote:

I don’t think the .so issue is caused by the swig fixes. Doing ‘make -j’
on master fails the same way.

But the swig doc issues that you were trying to fix do indeed appear to
have been fixed.

Both the above issues have been fixed and pushed to maint->master->next.

Johnathan