Hi Øyvind,
--- Øyvind Harboe <oyvind.har...@zylin.com> schrieb am Mi, 16.9.2009:
> This patch needs work.
>
> OpenOCD lacks the exception concept of a "try-catch".
> There
> are other cases where we *do* want error messages.
>
I agree that we need a message when it *finally* fails. However, when a
tentative call to alloc() can't be fulfilled, only the caller knows what that
means.
It may retry the allocation a few times with a smaller size, then give up after
a while. If this is fatal, it should issue a LOG_ERROR. If it can fall back to
another option, it may just silently drop the info. If that other option slows
things noticeably down, it may issue a LOG_INFO/LOG_WARNING.
I've just checked all calls to target_alloc_working_area().
With two exceptions, all modules do one of the following:
a) LOG_ERROR (ok)
b) LOG_WARNING (ok)
c) LOG_WARNING + fallback (ok)
d) LOG_INFO + fallback (ok)
e) no message, just fallback (ok)
f) LOG_DEBUG (arm_nandio,c) (a LOG_WARNING would be better)
Only target/armv7m.c and target/arm7_9_common.c just pass the error to the
caller. This is for the checksum and blank check algorithms.
I would suggest to add LOG_WARNING here.
I'm still convinced that the warning in target_alloc_working_area() is
misplaced. With the current situation, a bunch of warning may be issued, but
the software runs perfectly.
Regards,
Rolf
_______________________________________________
Openocd-development mailing list
Openocd-development@lists.berlios.de
https://lists.berlios.de/mailman/listinfo/openocd-development
