[dpdk-dev] [dpdk-ci] [PATCH] eal/windows: fix build warnings in MinGW

[Tal Shnaiderman]

> Subject: Re: [dpdk-ci] [PATCH] eal/windows: fix build warnings in MinGW
> 
> External email: Use caution opening links or attachments
> 
> 
> 2021-04-19 15:29 (UTC-0400), Aaron Conole:
> > Thomas Monjalon <thomas at monjalon.net> writes:
> >
> > > 19/04/2021 20:06, Tal Shnaiderman:
> > >> > > 2021-04-18 19:04 (UTC+0000), Tal Shnaiderman:
> > >> > > > > 18/04/2021 19:08, Tal Shnaiderman:
> > >> > > > > > the strncasecmp marco defined in rte_os_shim.h is already
> > >> > > > > > defined in MinGW-w64, as a result the compiler prints out
> > >> > > > > > the warning below on function redefinition whenever
> > >> > > > > > compiling a file including the
> > >> > > header.
> > >> > > > > >
> > >> > > > > > ..\lib/librte_eal/windows/include/rte_os_shim.h:21:
> > >> > > > > > warning: "strncasecmp" redefined #define strncasecmp(s1,
> > >> > > > > > s2,
> > >> > > > > > count) _strnicmp(s1, s2, count)
> > >> > > > >
> > >> > > > > Why the tests are passing? Are we allowing warnings in build
> test?
> > >> > > >
> > >> > > > It's a good question, I see CI is passing without warnings,
> > >> > > > Dmitry can you
> > >> > > reproduce it?
> > >> > > >
> > >> > > > On my side the collision is with the following file:
> > >> > > >
> > >> > > > C:/mingw-w64/x86_64/mingw64/x86_64-w64-
> > >> > > mingw32/include/string.h:119:
> > >> > > > note: this is the location of the previous definition
> > >> > > > #define strncasecmp _strnicmp
> > >> > > >
> > >> > > > I'm using:
> > >> > > > -gcc (x86_64-win32-seh-rev0, Built by MinGW-W64 project)
> > >> > > > 8.1.0 -meson
> > >> > > > 0.53.2
> > >> > >
> > >> > > I can't reproduce it using the same versions.
> > >> > > Do you invoke meson with other options than "-Dexamples=..."?
> > >> >
> > >> > This is the meson line I'm using:
> > >> >
> > >> > meson "-Dexamples=helloworld" build --wipe
> > >> > --default-library=static -- buildtype debug
> > >> >
> > >> > The problem however is that MinGW's string.h defines the macro in
> > >> > line 119, for some reason my build includes it and the rest doesn't.
> > >>
> > >> Hi Dmitry,
> > >>
> > >> I also noticed warnings in clang on my setup that are related to
> rte_os_shim.h:
> > >>
> > >> [185/314] Compiling C object
> drivers/a715181@@tmp_rte_common_mlx5 at sta/common_mlx5_mlx5_co
> mmon_mp.c.obj.
> > >> In file included from ../drivers/common/mlx5/mlx5_common_mp.c:13:
> > >> In file included from ..\drivers\common/mlx5/mlx5_common_utils.h:8:
> > >> In file included from ..\drivers\common/mlx5/mlx5_common.h:17:
> > >> ..\lib/librte_eal/windows/include\rte_os_shim.h:22:51: warning:
> > >> token pasting of ',' and __VA_ARGS__ is a GNU extension
> > >> [-Wgnu-zero-variadic-macro-arguments]
> > >> #define open(path, flags, ...) _open(path, flags, ##__VA_ARGS__)
> > >>                                                   ^ However don't
> > >> see it in CI, I'm using clang version 9.0.1
> > >
> > > It seems we should improve our CI.
> > > Please open suggestions in the CI bugzilla.
> >
> > Please do.
> >
> > Will this only be caught by mingw64 on windows?  Will we get the same
> > issues with a linux mingw install?  I'm guessing yes, but don't know
> > mingw very well.  We may be able to install the mingw package under
> > our github actions pipeline.
> 
> Yes, Linux MinGW-w64 produces the same warning (about strncasecmp).
> __VA_ARGS__ warning won't show up with GNU compiler, obviously.
> It may with ICC, but I don't have it to check.
> 
> I figured out it's not --default-library=static, but --buildtype=debug that
> caues the arning to appear, my mistake in previous message.

Right, using --buildtype=release doesn't reproduce the error.

BTW, how does release build passes if I excluded the definition from RTE_TOOLCHAIN_GCC?