On May 3 16:32, Reini Urban wrote: > 2011/5/3 Corinna Vinschen: > > I have a problem with the libsigsegv package. Right now the gawk > > maintainer started the testing cycle for the next gawk release, and the > > second time now the sigsegv.h header broke the build. > > > > The reason for the breakage is the same each time, sigsegv.h includes > > windows.h and thus clutters the namespace with stuff unexpected by a > > POSIX application: > > You are right. I'll try it out. > > > - The first time it was the definition of "WCHAR". > > > > - The second time it was the definition of "small"i, which, believe it > > or not, is defined as datatype in . > > > > - And there's still the definition of "min", which fortunately only > > produces a warning. > > > > AFAICS the only reason to include windows.h is to get the definition of > > the CONTEXT type which in turn is used to typedef stackoverflow_context_t. > > > > Is it really necessary to expose this to the application? Wouldn't it > > be better to define stackoverflow_context_t to void * for applications? > > THus we can drop the unfortunate inclusion of windows.h and not clutter > > the POSIX namespace with Windows definitions. > > Bruno will disapprove but for the cygwin port alone I can live with > that workaround.
Thanks, but why should Bruno not approve? Cygwin's task is to provide a POSIX environment. The fact that applications can still use Windows calls is a nice add-on, but non-Windows aware projects should have the right to get an environment which is not poluted with Windows definitions. Eric's patch should have gone in long ago. Corinna -- Corinna Vinschen Please, send mails regarding Cygwin to Cygwin Project Co-Leader cygwin AT cygwin DOT com Red Hat
