Hi Charlie, Harry & Gayarthi,
Would like to sort out soon error handling in ncap. The default error handling with ncap is to silently fail and then attempt to process the next line. Errrors can be seen with -D 2 option. Want to change behaviour - so genuine syntax errors and some run time errors cause program termination.
Whats ahappening is that because of this behaviour bm scripts with errors are being run. Also the behaviour is annoying when plugging leaks
The fix is gonna break a few bm & test scripts. If nobody has any objects I shall start doing the changes tomorrow
Regards Henry
If you would like to refer to this comment somewhere else in this project, copy and paste the following link:
Benchmark script breakage is fine I think as long as we know the reason. If there's a longer term change in the way the script handles errors. let me know what you'd like and I'll fix it from this end if it doesn't cause a cascade of catastrophe.
Harry
If you would like to refer to this comment somewhere else in this project, copy and paste the following link:
Sounds good. More stringent error handling should make ncap
more predictable. Regarding
> The fix is gonna break a few bm & test scripts.
It's fine to commit changes that break the regression testing
done by nco_bm.pl, as long as adequate notice and responsibility
for fixing them later are taken. It is not OK to commit changes
that break the benchmark part of nco_bm.pl. We need ncap
to always be able to handle the minimal requirements expected
of it (mainly LHS casting) in the benchmark scripts. So try to
keep the terminology straight when you mean regression
test and when you mean benchmark test. For clarification,
the file creation option to nco_bm.pl is part of the benchmarks.
Thanks,
Charlie
If you would like to refer to this comment somewhere else in this project, copy and paste the following link:
Hi Charlie, Harry & Gayarthi,
Would like to sort out soon error handling in ncap. The default error handling with ncap is to silently fail and then attempt to process the next line. Errrors can be seen with -D 2 option. Want to change behaviour - so genuine syntax errors and some run time errors cause program termination.
Whats ahappening is that because of this behaviour bm scripts with errors are being run. Also the behaviour is annoying when plugging leaks
The fix is gonna break a few bm & test scripts. If nobody has any objects I shall start doing the changes tomorrow
Regards Henry
Benchmark script breakage is fine I think as long as we know the reason. If there's a longer term change in the way the script handles errors. let me know what you'd like and I'll fix it from this end if it doesn't cause a cascade of catastrophe.
Harry
Hi Henry,
Sounds good. More stringent error handling should make ncap
more predictable. Regarding
> The fix is gonna break a few bm & test scripts.
It's fine to commit changes that break the regression testing
done by nco_bm.pl, as long as adequate notice and responsibility
for fixing them later are taken. It is not OK to commit changes
that break the benchmark part of nco_bm.pl. We need ncap
to always be able to handle the minimal requirements expected
of it (mainly LHS casting) in the benchmark scripts. So try to
keep the terminology straight when you mean regression
test and when you mean benchmark test. For clarification,
the file creation option to nco_bm.pl is part of the benchmarks.
Thanks,
Charlie