On Tue, 7 Mar 2023, LIU Hao wrote:
在 2023/3/7 20:16, Martin Storsjö 写道:
Right, thanks.
In
https://github.com/qemu/qemu/commit/dbd672c87f19949bb62bfb1fb3a97b9729fd7560
I got another wording suggestion - so maybe __USE_MINGW_SETJMP_WITHOUT_SEH
- does that seem reasonable to you?
Probably `__USE_MINGW_SETJMP_NON_SEH`... `WITHOUT_SEH` sounds like SEH was
something attached to setjmp that could be peeled off; it isn't. SEH is the
implementation and essence of the Microsoft `setjmp`.
Well in the case of the x86_64 version, the Microsoft implementation can
do a non-SEH traditional setjmp/longjmp too, by passing NULL for the
second parameter.
But I don't mind going with __USE_MINGW_SETJMP_NON_SEH either.
BTW would you mind examining the `USE_NO_MINGW_SETJMP_TWO_ARGS` macro too? It
may affect the prototype of `_setjmp()`, but despite that, there is really
only function to call, so how is the one-argument `_setjmp()` supposed to be
useful?
I guess I could have a look. My guess is that that codepath has bitrotted,
but maybe it still works on i386/x86_64 msvcrt.dll at least.
In case it's mostly broken and nobody is using it, what path forward do
you see for it? Just ripping it out?
// Martin
_______________________________________________
Mingw-w64-public mailing list
[email protected]
https://lists.sourceforge.net/lists/listinfo/mingw-w64-public
