Hi Hans, I used the official Ubuntu binary: http://llvm.org/releases/3.6.0/clang+llvm-3.6.0-x86_64-linux-gnu-ubuntu-14.04.tar.xz Seems to me that it should be an optional dependency. If the target system has tinfo or ncurses, it should use it, otherwise fallback to no color. On Mon, Mar 30, 2015 at 6:02 PM, Hans Wennborg <hans at chromium.org> wrote:> Hi Perry, > > On Mon, Mar 30, 2015 at 11:16 AM, Perry Hung <perry at leaflabs.com> wrote: > > I'm using the official 3.6 release binaries for x64 Linux. When compiling > > with warnings, I don't see any color diagnostics. Explicitly enabling > > -fcolor-diagnostics works fine. I have TERM set to xterm-256color. Any > idea > > why? > > Which of the release binaries are you using? 3.6 has x86_64 binaries > for Ubuntu, OpenSuSE and Fedora, and they were built by difference > people. > > From your ldd dump, it seems the binary you're using was built without > libtinfo, presumably because it wasn't installed on that machine and > we don't require it for building. > > - Hans >-------------- next part -------------- An HTML attachment was scrubbed... URL: <http://lists.llvm.org/pipermail/llvm-dev/attachments/20150330/2444bc8b/attachment.html>
On Mon, Mar 30, 2015 at 6:27 PM Perry Hung <perry at leaflabs.com> wrote:> Hi Hans, > > I used the official Ubuntu binary: > > http://llvm.org/releases/3.6.0/clang+llvm-3.6.0-x86_64-linux-gnu-ubuntu-14.04.tar.xz > > Seems to me that it should be an optional dependency. If the target system > has tinfo or ncurses, it should use it, otherwise fallback to no color. >There is no reasonable way to implement that IMO. It would require using dlopen to test for the library, open up a world of security and performance concerns, etc.> > On Mon, Mar 30, 2015 at 6:02 PM, Hans Wennborg <hans at chromium.org> wrote: > >> Hi Perry, >> >> On Mon, Mar 30, 2015 at 11:16 AM, Perry Hung <perry at leaflabs.com> wrote: >> > I'm using the official 3.6 release binaries for x64 Linux. When >> compiling >> > with warnings, I don't see any color diagnostics. Explicitly enabling >> > -fcolor-diagnostics works fine. I have TERM set to xterm-256color. Any >> idea >> > why? >> >> Which of the release binaries are you using? 3.6 has x86_64 binaries >> for Ubuntu, OpenSuSE and Fedora, and they were built by difference >> people. >> >> From your ldd dump, it seems the binary you're using was built without >> libtinfo, presumably because it wasn't installed on that machine and >> we don't require it for building. >> >> - Hans >> > > _______________________________________________ > LLVM Developers mailing list > LLVMdev at cs.uiuc.edu http://llvm.cs.uiuc.edu > http://lists.cs.uiuc.edu/mailman/listinfo/llvmdev >-------------- next part -------------- An HTML attachment was scrubbed... URL: <http://lists.llvm.org/pipermail/llvm-dev/attachments/20150330/1194ca22/attachment.html>
> > > There is no reasonable way to implement that IMO. It would require using > dlopen to test for the library, open up a world of security and performance > concerns, etc. > > I understand. I also understand servers may not wish to havencurses/tinfo, and it may be an onerous hard dependency. In any case, color diagnostics are advertised as default-on in the documentation. It seems reasonable to expect that the release binaries should reflect that. Is this worth filing a bug? In the meantime, I will bask in glorious color courtesy of trunk. -------------- next part -------------- An HTML attachment was scrubbed... URL: <http://lists.llvm.org/pipermail/llvm-dev/attachments/20150331/4c2f8cd2/attachment.html>