Hello,
On 11/10/2014 00:29, Alan Hourihane wrote:Yes, that's fine to do, and it's not strange. It is expected behaviour when you call Fselect(1,...) now.If I understand well, Olivier used Fselect(1,...) to estimate the time slice? It worked by chance, as a side effect of the previous implementation.
I use it to estimate time slice yes, but I have try to replace result by 0 to see the reaction of MyAES and it work fine, this is just a way to better adjust MU_TIMER answer, not more, as I know Fselect(1,...) is not forbiden.
So there is something else, not mix problems, ok I should not use Fselect(1,..) to know timeslice, but use Fselect(1,...) cause issue in MyAES absolutely not link to the timeslice measure. I have try to do program test but I'm not able to reproduce the issue for the moment, I should probably try debug in Mint itself, I have done in the past to be able to compil Mint with GCC4 while it compil only with GCC2.96.3, it is not a very good memory!
Not fine, ok MyAES work but there is possible other programs with problem, even I think this case is probably very uncommon. Imagine someone writing a new program and have the same issue how it is possible to link the issue with Fselect(1...)? I think he can search for month the problem! I update arround all years! If nobody prevent me very quickly I think I could search for a long time. There is perhaps a bug in MyAES (I don't think for this case) but we have not find explanation for the moment.Isn't there a better solution to know the time slice from user programs? Anyway, I still wonder why user programs could need that information.I'll put back the Fselect(1,...) fix in the kernel.Fine! So everything works again.
Olivier