Re: [PATCH] Use strwinerror in gdb/windows-nat.c

[Jon Turney <jon.turney@dronecode.org.uk>]

On 08/11/2022 18:24, Tom Tromey via Gdb-patches wrote:
>>>>>> "Jon" == Jon Turney <jon.turney@dronecode.org.uk> writes:
> 
> Jon> I'd suggest the attached to fix this, but I don't think this qualifies
> Jon> as obvious, since it's not obvious what USE_WIN32API is supposed to
> Jon> mean, so a review would be useful.
> 
> I don't fully know either, but this is what gdbsupport/common.m4 says:
> 
>    case ${host} in
>      *mingw32*)
>        AC_DEFINE(USE_WIN32API, 1,
> 		[Define if we should use the Windows API, instead of the
> 		 POSIX API.  On Windows, we use the Windows API when
> 		 building for MinGW, but the POSIX API when building
> 		 for Cygwin.])
>        WIN32APILIBS="-lws2_32"

Well, yes.  But this seems to mash up a few different things, e.g.:

- using the Windows API for interfacing with the debuggee
- not using any POSIX API
- target is MinGW

> Jon> Commit 02d04eac "Use strwinerror in gdb/windows-nat.c" also moves
> Jon> strwinerror() under the USE_WIN32API conditional, which is not defined
> Jon> for Cygwin (and looks like it shouldn't be, as appears to imply
> Jon> non-POSIX and MiNGW and WinSock...)
> 
> Jon> Also enable the declaration and definition of strwinerror() when
> Jon> __CYGWIN__ is defined.
> 
> I'm not sure how this area ought to work on Cygwin.  The goal here is to
> get a better error message when some Windows API fails.  If Cygwin can
> use strwinerror, then this seems fine.  If not, I could write a patch to

Thanks for reviewing this change.

Cygwin currently uses the Windows Debug API to control the inferior, so 
it's getting Windows error codes back, so turning those into text like 
this is indeed useful.

I'm going to take this as an approval and apply.

(There is a locale issue, in that the Windows idea of the current 
locale, which FormatMessage() uses, is not necessary the same, or even 
in a 1:1 mapping with Cygwin's POSIX-like idea of it, but I think I'll 
ignore that for the moment...)

> stub this out for Cygwin, which would be the status quo ante.

I think the status quo includes that these function would be used in 
gdbserver, if anyone built it for Cygwin, because there weren't under 
USE_WIN32API before.