Experimental Physics and
Industrial Control System

Benjamin Franksen <[email protected]> · Thu, 9 Jul 2015 21:07:57 +0200

I am writing to all interested parties as the current maintainer and 
main developer of the EPICS sequencer. The problem is:

(1) Seemingly harmless changes to the code organization or in the build 
system (i.e. Makefiles) have the tendency to break the build on Windows.

(2) Similarly, running the tests can also easily break on Windows due to 
differences in the way processes are started and stopped.

(3) Now and then I forget to avoid gcc'isms (mostly stuff that's not in 
the C90 standard) which break when trying to compile with native C 
compilers (like Visual Studio).

I have not done any serious work on Windows for years (perhaps decades). 
My original hope was that I would get some in-house support, i.e. people 
here at HZB who would set up the VMs I need for testing (Windows7 in 32 
and 64 bit, both with Mingw and native compiler), but that did not work 
out. Which means that, for now and in the forseeable future, I will not 
be able to test or even build on Windows, so I cannot rightfully claim 
to support it as a build and target platform for the sequencer. For a 
while Mark River has acted as de-facto co-maintainer (many thanks for 
that!) but his time is understandably limited, given the large number of 
projects he is responsible for.

This is why I am asking for a volunteer to act as a co-maintainer of the 
sequencer for the Windows side.

Your job would be to

(a) Notify me whenever building or running the tests breaks on Windows. 
Ideally this should be automated by extending the Jenkins build farm 
(which is currently maintained by Andrew Johnson) to cover the several 
Windows variants. But even with that, it doesn't make sense for me to 
publish patch after patch waiting hours in between until I find that I 
made a stupid error in a Perl script.

(b) Help me to react to reports from users about such problems: try to 
reproduce the problem, find out its cause, share your findings (or lack 
of them) with me and the user.

(c) Try to come up with a solution. Ideally send me a patch that solves 
it. Failing that, collaborate with me via email (or otherwise, perhaps 
meet in a chatroom) exchanging analysis and patches until we get it 
working again.

You should have Windows machines (VM or real) readily available and 
configured to allow building the sequencer and running the testsuite, 
ideally in (at least) the four variants I listed above. It would be 
perfect if you (or your colleagues) actually used the sequencer on 
Windows because some problems only come up in the real world and/or are 
hard to test automatically.

It is not necessary to be imtimately familiar with the code base itself, 
as I expect most problems to be of the sort described above: build 
failures and failures in the test system. I will of course help with 
getting started and will answer questions about the code, the build 
system, how the tests work etc, whenever they arise.

Being available on short notice is also not a requirement: the work can 
usually be postponed a week or two (or longer if necessary) until we 
both have enough time available. But at some point there must be the 
possibility to tightly work together for as long as it takes to fix 
things. (If we are in different time zones this period might stretch 
over a few days.) I expect the need for this kind of close collaboration 
to arise only occasionally, perhaps every few months, with long 
stretches of nothing (noticeable) happening in between.

If nobody volunteers I fear I'll have to remove any claim (in the 
documentation) that the sequencer supports Windows as a target or build 
platform, acknowlegding what has for a while been the state of things: 
if it breaks on Windows you are (more or less) on your own. Of course I 
will continue to apply patches people send me when they find a fix for a 
problem on Windows, provided it doesn't break anything on Linux.

Cheers
Ben
-- 
"Make it so they have to reboot after every typo." ― Scott Adams

Experimental Physics and Industrial Control System

Experimental Physics and
Industrial Control System