smeenai added inline comments.
================ Comment at: test/Driver/mingw-windowsapp.c:5-6 +// CHECK_DEFAULT: "-lmsvcrt" "-ladvapi32" "-lshell32" "-luser32" "-lkernel32" "-lmingw32" +// CHECK_WINDOWSAPP: "-lwindowsapp" "-lmingw32" +// CHECK_WINDOWSAPP-SAME: "-lmsvcrt" "-lmingw32" ---------------- mstorsjo wrote: > smeenai wrote: > > Why do we end up with -lmingw32 twice, and why not just check the full > > line, like you're doing for the default case? > I don't remember exactly why -lmingw32 ends up multiple times; I think it > comes from legacy compat with binutils ld, where the library ordering matters > more (a later static library doesn't trigger search in an earlier one). > > I'm checking twice, since there are other unrelated entries between these > that I didn't want to spell out, to avoid making the test overly specific (to > avoid having to update the test in case some of those are changed). Makes sense; I didn't realize before that all the other libraries in the default case were sandwiched between the -lmsvcrt and the -lmingw32 in the CHECK_WINDOWSAPP-SAME case. A -NOT check would perhaps have been more obvious, but I think that might end up interacting poorly with the other checks, so I'm fine with this as is. Repository: rC Clang https://reviews.llvm.org/D49059 _______________________________________________ cfe-commits mailing list cfe-commits@lists.llvm.org http://lists.llvm.org/cgi-bin/mailman/listinfo/cfe-commits
