Jeff Law <l...@redhat.com> wrote: >On 11/15/13 11:26, Tom Tromey wrote: >> PR bootstrap/58572 was inadvertently caused by the automatic >dependency >> patch series. >> >> The symptoms are that "make bootstrap-lean" will cause a subsequent >> "make install" to fail. >> >> The bug is that the automatic dependency code is picking up in-tree >> dependencies that were omitted by the old manual dependency tracking. >> You can see this by grepping for "libstdc" in $build/gcc/.deps/*. >> >> However, the "compare" step of "make bootstrap-lean" removes the >> previous build tree. This causes these dependencies to be >unresolved, >> and "make" thinks that the objects need to be rebuilt. This rebuild >is >> done with the system compiler, causing errors. >> >> This patch fixes the bug by removing the previous-stage-removal code >> from the top-level "compare" target. >> >> Tested by "make boostrap-clean; make install". >> >> Let me know what you think. >> >> Tom >> >> 2013-11-15 Tom Tromey <tro...@redhat.com> >> >> PR bootstrap/58572: >> * Makefile.in: Rebuild. >> * Makefile.tpl ([+compare-target+]): Don't remove previous stage. >But isn't the whole point of bootstrap-lean to remove those directories > >during the build?
If it reduces peak disk usage yes. These don't, no? That said, we could also 'fix' the dependencies at that point? Richard. >jeff
