On Tue, 29 Jan 2019 at 00:23, Shlomi Fish <shlomif@xxxxxxxxx> wrote: > > Hi! > > On Tue, Jan 29, 2019 at 1:50 AM Shlomi Fish <shlomif@xxxxxxxxx> wrote: >> >> Hi, >> >> On Tue, Jan 29, 2019 at 1:28 AM Jonathan Wakely <jwakely.gcc@xxxxxxxxx> wrote: >>> >>> On Mon, 28 Jan 2019 at 23:24, Shlomi Fish <shlomif@xxxxxxxxx> wrote: >>> > >>> > Hi all! >>> > >>> > I am trying to do a https://reproducible-builds.org/ setup in my C / gnu11 >>> > / cmake based FOSS project so I can cache valgrind results. The project is >>> > https://fc-solve.shlomifish.org/ and it has a git repo at >>> > https://github.com/shlomif/fc-solve . >>> >>> It's not clear from your mail if you're using some or all of GCC's >>> options that are relevant, -Wdate-time and -frandom-seed, and the env >>> var SOURCE_DATE_EPOCH. >> >> >> I am using -Wdate-time and -Werror and it also happens with export SOURCE_DATE_EPOCH="0" now that I tried it. Should I try -frandom-seed as well? >> > > After adding a hack of putting -frandom-seed=24 in the $CFLAGS, the problematic behaviour seems to be gone: > > https://github.com/shlomif/fc-solve/commit/836c3ed85c4d1ce5c2c700b93f284e6019f0c105 > > Thanks! Great. If that fixed it presumably the differences were in unique identifiers used for things like unnamed namespaces and lambda expressions.
