Experimental Physics and
Industrial Control System

Andrew Johnson via Core-talk <[email protected]> · Tue, 16 Jul 2019 17:45:31 -0000

Freddie, was your machine running Windows 7 or Windows 10? I have just disabled the retries in the unit test code and can't seem to trigger the failures any more (without adding the fflush()), although it may be that running the test program by hand doesn't do it. I probably saw them when I was using Windows 7 though, now I'm running Windows 10 and that might also have fixed the problem.

The opt_D flag is used by the EPICS build system to generate a make dependency rule for the output file. The expected behaviour is supposed to match GCC's -M flag and output the rule to stdout, using the -o argument to provide the name of the target filename to use in the rule. This code was a recent addition, and the author (most likely me) might not have considered the effect of an error that aborts the program. For it to call unlink(outFile) isn't a problem though, it ensures that GNUmake will have to run msi without the -D to (re)generate the output file, which *should* result in the same error and the build stopping.

I'm thinking we should add the flush(stdout), remove the retries from the test program and just wait to see whether we see any more failures.

-- 
https://code.launchpad.net/~freddie-akeroyd/epics-base/+git/epics-base/+merge/369763
Your team EPICS Core Developers is subscribed to branch epics-base:3.15.

Experimental Physics and Industrial Control System

Experimental Physics and
Industrial Control System