On Fri, Feb 03, 2017 at 09:13:23AM -0700, Martin Sebor wrote:
I'm not opposed to the changes, certainly not to cleaning things up,
but I don't think now is the time to be making these tweaks. Jakub's
patch is fine with me without those tweaks, and with the correction
What do you mean by my patch without those tweaks?
The intent of the patch is not just fix the diagnostics and
wrong-code issue, but primarily that any further needed fix will need
to tweak just a single spot, not many, otherwise e.g. you need to have
sufficient testsuite coverage for all those paths.
Say in the http://gcc.gnu.org/ml/gcc-patches/2017-02/msg00234.html
patch, you would with my patch need just the tree_digits part,
and then the very last hunk in gimple-ssa-sprintf.c (with
likely_adjust &&
removed). Because you do the adjustments only if !res.knownrange
and in that case you know argmin/argmax are actually dirtype's min/max,
so 0 must be in the range.
I'm not opposed to changing this to make it more intuitive or useful
but again, I would rather not spend time on it now. Instead, I would
prefer to discuss what we want after the dust from the release has
settled and implement a consistent solution in GCC 8.
I fear this isn't the last fix needed, the code is very complex and not
sufficiently covered by testcases, so if we don't want to make the code
more maintainable now, I'd be strongly for just turning
-fprintf-return-value off by default for the release. Bogus warnings
can be lived with or worked around, silent wrong-code is much worse.