Re: [FFADO-devel] FFADO-mixer and python

[Vladimir S. <sa...@gm...>]

Hello Jonathan!

I've finally solved the issue. But that was a huge cost to my nerves.
The development/maintenance team of SCons is inadequate and I strictly
recommend to drop scons build system from FFADO project.

Here's the final solution:
> I've found that regression occurred when changing from 3.0.4 version to 3.0.5. 
> I've inspected the changes
> in the SConf.py::_shutdown method and ONLY THEN found that the option 
> --config=force was the case.
> Removing it from build call solved the problem.

The overall discussion you may observe in this topic:
https://github.com/SCons/scons/issues/3579

That's the theater of non-constructive idiocy!

Best,
Vladimir

11.03.2020 12:28, Jonathan Woithe пишет:
> Hi Vladimir
>
> Further to my previous message, I just tried your build script on my
> development system.  It's running an older version of gcc so I had to remove
> "-fno-strict-aliasing-ggdb" from COMPILE_FLAGS.  Furthermore, the system
> python is python2, so that's what scons was running under.  Scons is v2.4.1
> and python is 2.7.17.
>
> With those notes, FFADO trunk (r2783) compiled successfully using your build
> script.  This makes it a little difficult for me to debug this issue
> further.
>
> The most obvious difference is the scons version.  Perhaps scons 3.1.2 is
> doing something that earlier versions did not do.
>
> I added your print statement to SConstruct in the same location as you:
>
>    print("!!!!!!!! %s\n" % (env.Dump())) # Added by me for debug
>
> Taking a look at the output, I see that it executes several times.  The
> first is in response to the "scons -c" command.  I can confirm that
> LIBRAW1394_FLAGS does not appear in that output.  However, I don't think
> it's expected to because a full SConstruct execution is not done when a
> clean operation is requested.  The absence of LIBRAW1394_FLAGS in this case
> certainly didn't trigger an error.
>
> The second execution reported many more items in the environment - as I'd
> expect, since this is the compilation pass.  In particular, I noted that
> LIBRAW1394_FLAGS was present (which explains why I had no errors).  Its
> value was that reported by pkg-config rather than the environment variable
> set in your build.sh, but the main point is that it was present.  This is
> clearly very different behaviour to what you see on your system.
>
> As to why the manual setting of LIBRAW1394_FLAGS isn't preserved, I think
> the respective conf.GetPKGFlags() probably overrides it.
>
> I am still very confused about why the script as it stands was working for
> David given your difficulties.  He was, I believe, using scons 3.1 with
> python3, which is very close to what's on your system.  It is therefore
> puzzling as to why you are seeing a problem.
>
> As before, any further insights you can obtain from your system would be
> helpful.
>
> Rest assured I would very much like to get to the bottom of what's going on
> here and get a fix in to trunk.
>
> Regards
>    jonathan