Hi Martin, On Monday 04 February 2008 03:39:56 Martin Pool wrote: > I agree, distcc should probably at least run these jobs locally, and > maybe print a message to tell you so. Im not so sure about rewriting > them.
Yes, rewriting them is probably too hard to do reliably. Simply forcing jobs with -march=native or -mtune=native to run locally is easy enough; I've added a patch that does just that, which I tested locally. Note that -mtune=native does not result in broken binaries, but it may result in code that is optimized for a different cpu than specified, so it's probably best to run jobs with this option locally as well. Kind regards, Maks Verver.
diff -ur distcc-2.18.3/src/arg.c distcc-modified/src/arg.c --- distcc-2.18.3/src/arg.c 2004-11-30 13:13:53.000000000 +0100 +++ distcc-modified/src/arg.c 2008-02-05 17:56:58.000000000 +0100 @@ -171,6 +171,12 @@ to distribute it even if we could. */ rs_trace("%s implies -E (maybe) and must be local", a); return EXIT_DISTCC_FAILED; + } else if (!strcmp(a, "-march=native")) { + rs_trace("-march=native generates code for local machine; must be local"); + return EXIT_DISTCC_FAILED; + } else if (!strcmp(a, "-mtune=native")) { + rs_trace("-mtune=native optimizes for local machine; must be local"); + return EXIT_DISTCC_FAILED; } else if (str_startswith("-Wa,", a)) { /* Look for assembler options that would produce output * files and must be local.
__ distcc mailing list http://distcc.samba.org/ To unsubscribe or change options: https://lists.samba.org/mailman/listinfo/distcc