Fix or silence warnings

[psiha]

currently we need to disable:
"-Wconversion"
"-Wdocumentation"
"-Wunused-variable"
before
#include <kwk/utility/container/shape.hpp>
#include <kwk/utility/linear_index.hpp>

[jfalcou]

o_O -Wdocumentation ??? Is that a clang specific flag ?

[psiha]

yes clang can detect broken doxygen documentation

[jfalcou]

Good ! I didn't know it did that.
We'll try to see why the two other are not picked up by Wall Wother

[psiha]

for unused-variable you need to build release/without asserts (as some variables you use only for sanity checks)

[jfalcou]

Will add a CI recipe for at least one Release mode test then.

[psiha]

Will add a CI recipe for at least one Release mode test then.

you didn't have that already?? 🙈 🙈 :P

we'd prefer you test a full -Ofast -flto for the release build :)

[jfalcou]

Will add a CI recipe for at least one Release mode test then.
you didn't have that already?? 🙈 🙈 :P

I ported the CI script from EVE without rechecking them :p

we'd prefer you test a full -Ofast -flto for the release build :)

Can do

[jfalcou]

-Wunused and -Wdocumentation has been fixed

-Wconversion is another beast. It complains when passing int to [] that 1.f+i can't be assigned to float in tests.
It also makes all our design of size beign signed go to the drain cause we need to add tons of cast when interacting
with some std components.

Not sure how to handle those :/

[jfalcou]

OK found the one culprit, next PR will fix those

[psiha]

-Wunused and -Wdocumentation has been fixed

-Wconversion is another beast. It complains when passing int to [] that 1.f+i can't be assigned to float in tests. It also makes all our design of size beign signed go to the drain cause we need to add tons of cast when interacting with some std components.

Not sure how to handle those :/

You chose a signed size_type for vectorization-performance issues, avoiding exactly these warning issues and/or something else?

[jfalcou]

In fact the main problem was the way we were capturing fixed size. Went back to fixed size being unsigned when built from _c and silenced warnings in a few places with pragma. Other fixes were stupid mistake and adapting tests.

This have been merged.
Tell me if it is enough.

[jfalcou]

Closingfor now as multiple runs show no extra warning on our side

[psiha]

kiwaku/include/kwk/detail/sequence/prefilled.hpp:65:81: error: non-constant-expression cannot be narrowed from type 'unsigned long long' to 'unsigned long' in initializer list [-Wc++11-narrowing]

                                return std::array<std::size_t, static_size>{m...};

while building for emscripten (which is a 32bit platform so that could be difference)...

[jfalcou]

Oh I forgot about emscriptem.
Will fix this missign static_cast<> I guess then add an emscriptem target for CI

[psiha]

Oh I forgot about emscriptem. Will fix this missign static_cast<> I guess then add an emscriptem target for CI

Perhaps rather do not use 64bit types unconditionally - probably a size_t-like type would do in most those situations...
Also, even using explicitly 32bit types (instead of size_t) can give smaller codegen (if there is no mixing with 64bit types) when you know that you don't need the range (e.g. for 'certainly tiny' numbers like ranks/numbers of dimensions)..

[jfalcou]

Perhaps rather do not use 64bit types unconditionally - probably a size_t-like type would do in most those situations... Also, even using explicitly 32bit types (instead of size_t) can give smaller codegen (if there is no mixing with 64bit types) when you know that you don't need the range (e.g. for 'certainly tiny' numbers like ranks/numbers of dimensions)..

std::size_t is supposed to be the size of the system memory. It is not 64 bits ona 32 bits system.