On 12 April 2016 at 13:00, Jonathan Wakely <jwakely.gcc@xxxxxxxxx> wrote: > On 31 March 2016 at 19:30, onkel.jack@xxxxxxxxxxx > <onkel.jack@xxxxxxxxxxx> wrote: >> >> Thanks Kai. >> >> In the meantime I did some more experiments after I figured out I can repro the problem on RHEL6 too which does build much faster ... >> >> It turned out, the problem goes away if newlib gets configured with --enable-newlib-multithread, I lost that, for whatever reason. >> Somehow it seems to have an influence on configure libstdc++. Guess a diff in newlib headers might find it. >> I think, threading in gcc and newlib needs to be configured consistent despite the "single" default means NO, maybe its different in libstdc++ ... > > single means NO, but single is not necessarily the default! > > The docs you quoted say: > > "Beware that on some systems, GCC has not been taught what threading > models are generally available for the system. In this case, > --enable-threads is an alias for --enable-threads=single." > > But that's not the case for your target. GCC has been taught what > threading models are available for arm-geabi using newlib, so when you > say --enable-threads it assumes you want to enable threads! Oh sorry, I see it was Kai who quoted those docs. But anyway, even if you haven't taught libstdc++ about your target it might be guessing from newlib, and I think this is still relevant: > If you want to disable threads then use --disable-threads. Using > --enable-threads to disable threads based on your assumption about a > default is just silly.
