In the meantime I've committed my version of the patch to the gccgo branch. It will be updated to whatever the final mainline version is the next time I merge from mainline to the gccgo branch.
Ian On Thu, Nov 7, 2013 at 10:16 AM, Ian Lance Taylor <i...@google.com> wrote: > On Thu, Nov 7, 2013 at 8:48 AM, Bruce Korb <bk...@gnu.org> wrote: >> >> This time, I'm on my dev box and looked at the code. >> You remembered correctly that the first file name in the list >> of file names needs to not have wild card characters so that >> the testing scheme can create a file by that name. A directory >> named "*" is possible, but inconvenient. >> >> Anyway, it also matches prior art in: >> >>> /* >>> * glibc_c99_inline_3 >>> */ >>> fix = { >>> hackname = glibc_c99_inline_3; >>> files = bits/string2.h, '*/bits/string2.h'; >> >> >> but is inconsistent with: >> >>> /* Some versions of glibc have a version of bits/string2.h that >>> produces "value computed is not used" warnings from strncpy; fix >>> this definition by using __builtin_strncpy instead as in newer >>> versions. */ >>> fix = { >>> hackname = glibc_strncpy; >>> files = bits/string2.h; >> >> >> Hmmm. Does that fix need fixing, too? > > I didn't worry about that and the other cases because they are fixing > problems that existed in past glibc versions before Ubuntu changed > their directory layouts. But of course there is nothing wrong with > fixing them too. > > Your proposed patch looks fine to me and I think you should go ahead > and commit. Or I can if you prefer for some reason. > > Ian