RE: BETA: ci-scripts 3.0 (feat. Windows on Travis)

[Mark Rivers via Core-talk <core-talk at aps.anl.gov>]

Hi Ralph,

 

Mea culpa, I should have looked at the README.md more carefully. ADD_MODULES was exactly what I was looking for.

 

Ø  Yes, without the extension.

Ø  (Documented in the README.)

 

Actually I think the README only documents omitting the .set in the context of the “include” statement, not the SET statement.  There is one example (SET=stable) that omits the .set.

 

Ø  MinGW builds are now using the 'compiler: gcc' setting, both for the cross builds from Linux and for the native Windows builds.

Ø  You can just remove the 'compiler:' setting from your job lines, as 'gcc' is the default.

 

Removing compiler: mingw fixed the error, the mingw jobs now succeed.

 

Ø  Also: the 'STATIC=YES/NO' setting has been replaced by 'BCFG=default/static/debug/static-debug'.

 

Yes, I had made all of those changes already.

 

Now my asyn build is only failing on the static-debug builds on Windows with both gcc and vs2017.  https://travis-ci.org/github/epics-modules/asyn/builds/696901014

Those builds fail for different reasons.

 

Michael said those are known problems that are under investigation?

 

The gcc build is just “timing out” with no response after 10 minutes.  I sometimes see a similar problem on my Windows machine if I use “make –j” rather than “make –j8” for example.  One “make” process is just sitting there using 100% of a core.  What –j count is being used by that job?  Is see that cue.py has these lines:

 

cue.py:    parallel = kws.pop('parallel', 2)

cue.py:        makeargs = ['-j{0}'.format(parallel)]

 

I suspect that means there is a default value of –j2.  How does one specify a different value for “parallel”?

 

Thanks,

Mark

 

 

 

From: Core-talk <core-talk-bounces at aps.anl.gov> On Behalf Of Ralph Lange via Core-talk
Sent: Wednesday, June 10, 2020 5:46 AM
To: EPICS Core Talk <core-talk at aps.anl.gov>
Subject: Re: BETA: ci-scripts 3.0 (feat. Windows on Travis)

 

Hi Mark,

 

On Wed, 10 Jun 2020 at 02:00, Mark Rivers <rivers at cars.uchicago.edu> wrote:

 

I have a question about what version of EPICS base BASE=7.0 actually is.  Is that a specific tag, or is it the tip of the 7.0 branch?

 

As Michael wrote, the setting must refer to the name of a branch or tag. In the case of the EPICS Base repo, '7.0' is the name of a branch.

 

Another question.  In base7.set I define MODULES="sncseq".  When building for RTEMS I use base7.set, but I want to add “ipac” to MODULES.  Is there a syntax to do that?  If I use this line:

  - env: SET=base7 RTEMS=4.10 MODULES="ipac"

 

There is the ADD_MODULES setting to allow adding modules to specific builds.

(Documented in the README.)

 

First I had a minor problem.  The .travis.yml.example-full file does not have an example of using the SET argument.  I was doing SET=base7.set, and I got an error message saying .ci-local/base7.set does not exist. But it does exist.  The problem is that the .set extension must be omitted, so it should be SET=base7.

 

Yes, without the extension.

(Documented in the README.)

I will fix the misleading error message.

 

Some of the builds are working, but some are failing for reasons I don’t understand. You can see them here: https://travis-ci.org/github/epics-modules/asyn/builds/696643004

 

The mingw cross builds for Windows  both fail with this error:

 

$ mingw --version

Traceback (most recent call last):

[...]

 

MinGW builds are now using the 'compiler: gcc' setting, both for the cross builds from Linux and for the native Windows builds.

You can just remove the 'compiler:' setting from your job lines, as 'gcc' is the default.

(Documented in the example config file.) 

 

Also: the 'STATIC=YES/NO' setting has been replaced by 'BCFG=default/static/debug/static-debug'.

(Also shown in the example config file.)

 

Cheers,
~Ralph