Thanks for the diffs. It's surprising to me that there are any diffs
at all. I don't see any problematic differences, though.
** [issues:#5649] flower: Add boolean return value to 'Rational' class.**
Status: Started Created: Sun Dec 29, 2019 10:08 AM UTC by Werner LEMBERG Last Updated: Mon Dec 30, 2019 05:06 AM UTC Owner: Werner LEMBERG Attachments:
I think it would still be good to know how they come about. Maybe we
get zero values by having the numerator rather than the sign zero? If
so, it would be important to figure out where.
I think it would still be good to know how they come about.
Maybe we get zero values by having the numerator rather
than the sign zero? If so, it would be important to figure out
where.
Attached is a log file of
make EXTRA_CXXFLAGS="-W -Wall -Wconversion -Wno-sign-conversion"
for commit afb36349 (i.e., without my patch) that shows the warning locations.
I have not yet found a good reason for the regtest changes. Let me try some more experiments. Put to waiting, but I'll not hold it up more than a few days.
If you would like to refer to this comment somewhere else in this project, copy and paste the following link:
I am not able to trigger a difference in one session adding debug code. I am going to try the full make check thing to see whether this can make a difference.
If you would like to refer to this comment somewhere else in this project, copy and paste the following link:
What's really irritating is that touching flower/include/rational.hh does not trigger any recompile. Has this always been the case or is this a recently introduced dependency problem?
If you would like to refer to this comment somewhere else in this project, copy and paste the following link:
Sorry, sometimes the SourceForge user interface sucks. I set this to "Patch-push" again after not being able to corroborate any changes with either GCC and Clang, and the dependency problem occurs only with Clang which appears to fail at autogenerating dependencies, so you always have to do make clean before getting anything of consequence done. Which sucks, but is not a priority for our normal GCC-based process.
If you would like to refer to this comment somewhere else in this project, copy and paste the following link:
make check does not show any difference to the baseline without the patch. While this difference might depend on the compiler version, the current gcc version I use should be comparatively relevant. I have not been able to trigger any difference with either our code base or hand-crafted test programs either.
So I lean towards considering this a fluke, possibly because of a non-up-to-date test-baseline or some compiler bug? Setting back to Patch-push.
If you would like to refer to this comment somewhere else in this project, copy and paste the following link:
make check does not show any difference to the baseline without the patch. While this difference might depend on the compiler version, the current gcc version I use should be comparatively relevant. I have not been able to trigger any difference with either our code base or hand-crafted test programs either.
So I lean towards considering this a fluke, possibly because of a non-up-to-date test-baseline or some compiler bug? Setting back to Patch-push.
If you would like to refer to this comment somewhere else in this project, copy and paste the following link:
Passes make, make check and a full make doc.
Reg test diffs attached
Thanks for the diffs. It's surprising to me that there are any diffs at all. I don't see any problematic differences, though.
"Werner LEMBERG" wlemb@users.sourceforge.net writes:
I think it would still be good to know how they come about. Maybe we
get zero values by having the numerator rather than the sign zero? If
so, it would be important to figure out where.
--
David Kastrup
Related
Issues:
#5649Attached is a log file of
for commit afb36349 (i.e., without my patch) that shows the warning locations.
Which clang version?
clang++ 9.0 on MacOS Lion.
Leaving on Review for now.
Moving to countdown for Jan 5th - if you want to change this (because of the reg test diffs) feel free.
This patch has been counted down to push. If you want to change this (because of the reg test diffs) feel free.
I have not yet found a good reason for the regtest changes. Let me try some more experiments. Put to waiting, but I'll not hold it up more than a few days.
Thanks for working on this!
I am not able to trigger a difference in one session adding debug code. I am going to try the full
make checkthing to see whether this can make a difference.What's really irritating is that touching
flower/include/rational.hhdoes not trigger any recompile. Has this always been the case or is this a recently introduced dependency problem?I'm not aware of fundamental changes in this regard. Maybe Dan Eble knows more.
Sorry, sometimes the SourceForge user interface sucks. I set this to "Patch-push" again after not being able to corroborate any changes with either GCC and Clang, and the dependency problem occurs only with Clang which appears to fail at autogenerating dependencies, so you always have to do
make cleanbefore getting anything of consequence done. Which sucks, but is not a priority for our normal GCC-based process.make checkdoes not show any difference to the baseline without the patch. While this difference might depend on the compiler version, the current gcc version I use should be comparatively relevant. I have not been able to trigger any difference with either our code base or hand-crafted test programs either.So I lean towards considering this a fluke, possibly because of a non-up-to-date test-baseline or some compiler bug? Setting back to Patch-push.
Crikey. All of my testing used Clang. Back to waiting while I redo checks (but after the current other problem I am tracking)
Nope, still clean, no regtest differences, no behavior change when debug code checks explicitly.
make checkdoes not show any difference to the baseline without the patch. While this difference might depend on the compiler version, the current gcc version I use should be comparatively relevant. I have not been able to trigger any difference with either our code base or hand-crafted test programs either.So I lean towards considering this a fluke, possibly because of a non-up-to-date test-baseline or some compiler bug? Setting back to Patch-push.