Using SEH as the default on x86_64 makes sense, it is 10x more popular than SJLJ with mingw-w64
https://sourceforge.net/projects/mingw-w64/files/Toolchains%20targetting%20Win64/Personal%20Builds/mingw-builds/4.9.2/threads-win32/ Thanks! 2014-11-14 4:07 GMT+02:00 Reid Kleckner <[email protected]>: > I finally went ahead and removed -fseh-exceptions and made that behavior > the default on x86_64-*-windows-* in r221963. Eventually I intend to change > the EH personality functions for *-*-windows-msvc, but that shouldn't > affect mingw64. OK with you? > > On Wed, Sep 17, 2014 at 9:05 PM, Yaron Keren <[email protected]> > wrote: > >> I have not seen any MinGW distribution using DWARF exceptions on x64. >> They all use SEH or SJLJ. >> SEH should be the default unless -fsjlj-exceptions. >> >> Yaron >> >> >> 2014-09-18 4:08 GMT+03:00 Reid Kleckner <[email protected]>: >> >>> On further reflection, I think the nomenclature here is wrong. We need >>> to try harder to disambiguate two concepts of SEH: the __try / __except >>> constructs, and the Windows unwind and exception info format on x64, arm, >>> and probably ppc. In LLVM MC, we currently use the terminology WinCFI and >>> WinEH to refer to essentially the stuff in .pdata and .xdata respectively. >>> >>> -fseh-exceptions and LangOpts.SEHExceptions are poorly named because the >>> reader might assume that they enable an implementation of __try / __except, >>> which is incorrect. Can you help me come up with a less confusing name? My >>> strawmen are: >>> -fpdata-exceptions / PDataExceptions >>> -fxdata-exceptions / XDataExceptions >>> -fwin-eh-exceptions / WinEHExceptions >>> -fwin-cfi-exceptions / WinCFIExceptions >>> >>> These names are meant to stand as alternative to -fsjlj-exceptions and >>> dwarf cfi, the default. >>> >>> Alternatively, is this option even useful? Would you be OK with >>> completely removing this option in favor of only supporting .pdata >>> exceptions and sjlj exceptions on mingw64 x64? I noticed that the >>> mingwbuilds installer doesn't even present dwarf exceptions as an option >>> anymore (did it ever?). >>> >>
_______________________________________________ cfe-commits mailing list [email protected] http://lists.cs.uiuc.edu/mailman/listinfo/cfe-commits
