Re: [PATCH trvivial] linux-user/syscall.c: replace function pointers for flock64 fcntl with macros

[Michael Tokarev]

08.09.2024 20:39, Richard Henderson wrote:
> On 9/8/24 00:16, Michael Tokarev wrote:
> > For 32bit ARM fcntl64 syscall there are 2 possible argument types,
> > depending on cpu_env->eabi.  For other architectures, it is plain
> > struct flock64 in all cases.  In order to solve this, old code
> > used to take address of the conversion function and and run it
> > through this pointer.  Instead, introduce 2 helper macros for
> > the flock64 conversion, and define them in the same block where
> > the oabi conversion functions are defined, and use these
> > helpers directly in the actual code, making each part more
> > self-contained and easier to read.
> >
> > Also add comment to the block of code which itroduces the oabi
> > conversion functions.
> >
> > Note also there was an inconsistency in the old code: different
> > the differences in single place we eliminate this difference.
> >
> > While at it, replace tabs with spaces in nearby code.
> >
> > Signed-off-by: Michael Tokarev<mjt@tls.msk.ru>
> > ---
> >   linux-user/syscall.c | 40 +++++++++++++++++++++++-----------------
> >   1 file changed, 23 insertions(+), 17 deletions(-)
>
> Why do you think this is an improvement?

It just feels more natural, so to say.

> What was wrong with the function pointers?

Not exactly wrong.  It just hurts my eyes when I see an address
is taken of a function marked `inline` (though I understand well
this keyword is just a hint and the compiler is free to omit
inlining).  Also the typedefs are a bit ugly.

I come through this place when removing remnants of LFS, and
disliked this way.

Thanks,

/mjt