When compiling GCC 9.2.0 (which has failed many times for me now), I notice that stage 2 build GCC with debug info: # cd /var/tmp/portage/sys-devel/gcc-9.2.0-r2/work/build # file prev-gcc/xg++ gcc/xg++ prev-gcc/xg++: ELF 64-bit LSB executable, x86-64, version 1 (SYSV), dynamically linked, interpreter /lib64/ld-linux-x86-64.so.2, for GNU/Linux 3.2.0, not stripped gcc/xg++: ELF 64-bit LSB executable, x86-64, version 1 (SYSV), dynamically linked, interpreter /lib64/ld-linux-x86-64.so.2, for GNU/Linux 3.2.0, with debug_info, not stripped This uses a lot of extra disk space while building and probably also more CPU / time. # du -hs * [...] 1,4G gcc [...] 352M prev-gcc I just wonder if this is intended behavior and if so, why? Reproducible: Always
By default gcc builds itself 3 times: - stage1: build gcc-9.2.0 from any other system C++ compiler, usually without optmisations - stage2: build gcc-9.2.0 using -stage1 - stage3: build gcc-9.2.0 using -gtoggle and stage2 then gcc strips stage2/stage3 and makes sure the final result is identical. Build fails if it does not. stage2/3 are especially useful for USE="lto" and USE="pgo" phases where optimisations are performed by a version of a compiler being built. If you use none of the above you can use EXTRA_ECONF=--disable-bootstrap. Then gcc will use stage1/stage2 only to get final result without a comparison.
@Sergei: First, thank you for taking the time to answer. It's still unclear to me why stage2 and stage3 is compiled with debug info. Maybe it was there in your answer, but I didn't get it. (A little googling reveals that from time to time, the stage comparison fails because of debugging being enable.) Slightly hijacking my own issue now: My compilation of GCC 9.2.0 once again failed (on my industrial PC), this time I got all the way to the actual stage comparison and got "Bootstrap comparison failure!". So I quickly tested your suggestion to disable bootstrap: # EXTRA_ECONF="--disable-bootstrap" emerge -av gcc:9.2.0 [...] >>> Source configured. >>> Compiling source in /var/tmp/portage/sys-devel/gcc-9.2.0-r2/work/gcc-9.2.0 ... * Compiling gcc (bootstrap-lean)... make -j4 'LDFLAGS=-Wl,-O1 -Wl,--as-needed' STAGE1_CFLAGS= LIBPATH=/usr/lib/gcc/x86_64-pc-linux-gnu/9.2.0 'BOOT_CFLAGS=-m64 -O2 -march=bonnell -mtune=generic -fno-stack-protector' bootstrap-lean make: *** No rule to make target 'bootstrap-lean'. Stop. It failed immediately. By the way, I tested this on my workstation too, and it failed there as well. Let me know if I should file a bug report for this.
(In reply to Hans F. Nordhaug from comment #2) > @Sergei: First, thank you for taking the time to answer. > > It's still unclear to me why stage2 and stage3 is compiled with debug info. > Maybe it was there in your answer, but I didn't get it. (A little googling > reveals that from time to time, the stage comparison fails because of > debugging being enable.) Only one of two is compiler with debugging info: only stage2 or only stage3 (which of those depends on users' CFLAGS value). The reason gcc does it is to make sure gcc's code generation is stable and has no non-determinism from build to build. > Slightly hijacking my own issue now: > > My compilation of GCC 9.2.0 once again failed (on my industrial PC), this > time I got all the way to the actual stage comparison and got "Bootstrap > comparison failure!". That should not happen. Please file a bug. > So I quickly tested your suggestion to disable > bootstrap: > > # EXTRA_ECONF="--disable-bootstrap" emerge -av gcc:9.2.0 > [...] > >>> Source configured. > >>> Compiling source in /var/tmp/portage/sys-devel/gcc-9.2.0-r2/work/gcc-9.2.0 ... > * Compiling gcc (bootstrap-lean)... > make -j4 'LDFLAGS=-Wl,-O1 -Wl,--as-needed' STAGE1_CFLAGS= > LIBPATH=/usr/lib/gcc/x86_64-pc-linux-gnu/9.2.0 'BOOT_CFLAGS=-m64 -O2 > -march=bonnell -mtune=generic -fno-stack-protector' bootstrap-lean > make: *** No rule to make target 'bootstrap-lean'. Stop. > > It failed immediately. By the way, I tested this on my workstation too, and > it failed there as well. Let me know if I should file a bug report for this. You will also need to override 'bootstrap-lean' target with GCC_MAKE_TARGET=all. I'm using the following env file occasionally to skip some sanity checks: $ cat /etc/portage/env/gcc-nobootstrap.conf # skip unoptimised build EXTRA_ECONF=--disable-bootstrap GCC_MAKE_TARGET=all
Thanks again for a very useful reply. I'm marking the issue as resolved / fixed. And thanks for the gcc-nobootstrap env tip. However, I didn't need it this time. The compile actually completed last night after at least 10 tries. Probably just bad hardware - running Gentoo from a SD card.
