Running the regression tests on an embedded device
bodrato at mail.dm.unipi.it
Sat Jan 2 18:42:12 UTC 2016
Il Sab, 26 Dicembre 2015 2:14 am, Mike Mohr ha scritto:
> all tests must be compiled into a shared object. To facilitate this
> test suite, some changes must be made to the individual test programs:
> 1) All calls to exit and abort must be replaced with something else.
> 2) The function name "main" would have to be replaced by something else,
> 3) Any test-specific data located in the source files would have to
> be explicitly made static.
Also local functions should be made static, we have plenty of check_one,
But the main issue I see involves the global state.
If a test fails we will surely have memory leaks and possibly other problems.
Even if a test succeeds, it may impact on the status left for the next
In my opinion you can obtain a single executable collecting many tests,
but not all of them. Rewriting and maintaining all of the tests so that
none of them have any "side effect" is not worth the effort, I guess.
> If I were to do so, would the GMP developers be amenable
> to receiving and applying a patch to the test suite?
Do you have a general target in mind that can work for any
(cross-)platform compilation? Something like 'make check-executables'?
Should it allow both static and shared builds?
More information about the gmp-discuss