On 2010-03-26 21:07, Ralf Wildenhues wrote: > tags +upstream > thanks > > Hello Török, > > thanks for the report. > > * Török Edwin wrote on Thu, Mar 25, 2010 at 10:24:08PM CET: >> Currently libtool strips -flto when called for linking which is wrong. >> It should pass -flto to gcc, since its both a compile-time and a link-time >> flag. > > I agree. There has been a similar request for another compiler > recently, too, let's address them both, upstream.
I'm not subscribed to the upstream ML, but when you have any patches you'd like me to test, please send me an e-mail. > >> Actually I don't know why libtool strips any flags at all. If I pass a flag >> at compile time, and libtool happily passes it along (such as -flto here), >> why does it strip it at link time? > > FWIW, libtool only strips flags when creating shared libraries, not when > creating programs. Didn't know that, since I didn't get that far (build stopped when shared lib failed to build). Good to know though. > > Many flags need changed libtool semantics, when passed through. > For example, just passing through -flto without further changes > let me find one GCC bug Maybe you hit this bug? (mixing -O flags) http://gcc.gnu.org/bugzilla/show_bug.cgi?id=42402 > and half a dozen Libtool testsuite failures > and several places where libtool needs to be changed. It's simply > more robust to strip unknown flags. Fair enough. Could libtool give a warning when it strips a flag though? Not everyone inspects the link command-line passed to gcc, so they may think they used an optimization flag (other than -O), but in reality libtool stripped it. Best regards, --Edwin -- To UNSUBSCRIBE, email to debian-bugs-dist-requ...@lists.debian.org with a subject of "unsubscribe". Trouble? Contact listmas...@lists.debian.org
