On Tue, Mar 05, 2019 at 11:10:20AM -0600, Segher Boessenkool wrote: > On Tue, Mar 05, 2019 at 04:22:17PM +0100, Jakub Jelinek wrote: > > powerpc-linux-gnu is apparently the only target that provides > > MULTIARCH_DIRNAME unconditionally, > > (Unless building with default CPU a 4xx or 8xx or similar embedded. Which > doesn't seem right, oh well). > > > all others properly wrap that with > > if_multiarch, which decides if it should be used (--enable-multiarch, > > or if the test for automatic multiarch succeeds), or should be empty > > (--disable-multiarch, or the auto test failed). > > So what behaviour changes with this? What bug does it fix?
It changes multiarch_dir variable in the driver, which has various consequencies. Not that familiar with multiarch to know what exactly. PR89587. From my POV of view, the patch brings in consistency. > > Fixed thusly, bootstrapped/regtested on powerpc64{,le}-linux, also > > tested without and with this patch in powerpc-linux non-bootstrap and > > testing what -print-multiarch prints. Ok for trunk? > > Since Matthias thinks it is okay: it is okay :-) Thanks (to you both)! Thanks. Jakub