147298 2006-09-12 05:04 0000 [GCC4] app-doc/doxygen broken when compiled w/ -O3 2006-09-14 19:36:47 0000 1 1 1 Unclassified Gentoo Linux Applications unspecified x86 Linux RESOLVED FIXED http://forums.gentoo.org/viewtopic-t-495600.html P2 normal --- 1 chat.noir@arcor.de nerdboy@gentoo.org dev-tools@gentoo.org chat.noir@arcor.de 2006-09-12 05:04:25 0000 Compiling doxygen with GCC 4 and -O3 causes doxygen to behave improperly. Could verify it myself with app-doc/doxygen-1.4.[347], gcc-4.1.1 and CFLAGS="-march=athlon-xp -O3 -mfpmath=sse -msse". Setting optimization level down to -O2 made the problem disappear, so it might be a problem with GCC. Still the ebuild might want to check for affected versions of gcc and perform some actions to avoid the problem, or whatever is more suitable. chat.noir@arcor.de 2006-09-12 05:06:02 0000 Created an attachment (id=96784) Output of emerge info nerdboy@gentoo.org 2006-09-14 18:34:19 0000 Yes, -O3 still produces incorrect optimizations. In general, it's best not to try to over-optimze code with gcc; not just because -O3 is buggy, but also because in the majority of cases, gcc is way smarter about it. It will automatically enable the stuff that works, including mmx, sse, etc, when it can. Forcing things OTOH, breaks packages more often than not. I would just use -O2, along with a few other safe things, and not worry about it... nerdboy@gentoo.org 2006-09-14 19:36:47 0000 Checking and filtering... 96784 2006-09-12 05:06 0000 Output of emerge info emerge-info text/plain UG9ydGFnZSAyLjEuMSAoZGVmYXVsdC1saW51eC94ODYvMjAwNS4xLCBnY2MtNC4xLjEsIGdsaWJj LTIuNC1yMywgMi42LjE3LjYgaTY4NikKPT09PT09PT09PT09PT09PT09PT09PT09PT09PT09PT09 PT09PT09PT09PT09PT09PT09PT09PT09PT09PT09PT0KU3lzdGVtIHVuYW1lOiAyLjYuMTcuNiBp Njg2IEFNRCBBdGhsb24odG0pIFhQIDE4MDArCkdlbnRvbyBCYXNlIFN5c3RlbSB2ZXJzaW9uIDEu MTIuNApMYXN0IFN5bmM6IFR1ZSwgMTIgU2VwIDIwMDYgMDg6NTA6MDEgKzAwMDAKYXBwLWFkbWlu L2VzZWxlY3QtY29tcGlsZXI6IFtOb3QgUHJlc2VudF0KZGV2LWxhbmcvcHl0aG9uOiAgICAgMi4z LjUsIDIuNC4zLXIxCmRldi1weXRob24vcHljcnlwdG86IDIuMC4xLXI1CmRldi11dGlsL2NjYWNo ZTogICAgIFtOb3QgUHJlc2VudF0KZGV2LXV0aWwvY29uZmNhY2hlOiAgW05vdCBQcmVzZW50XQpz eXMtYXBwcy9zYW5kYm94OiAgICAxLjIuMTcKc3lzLWRldmVsL2F1dG9jb25mOiAgMi4xMywgMi41 OS1yNwpzeXMtZGV2ZWwvYXV0b21ha2U6ICAxLjRfcDYsIDEuNSwgMS42LjMsIDEuNy45LXIxLCAx LjguNS1yMywgMS45LjYtcjIKc3lzLWRldmVsL2JpbnV0aWxzOiAgMi4xNi4xLXIzCnN5cy1kZXZl bC9nY2MtY29uZmlnOiAxLjMuMTMtcjMKc3lzLWRldmVsL2xpYnRvb2w6ICAgMS41LjIyCnZpcnR1 YWwvb3MtaGVhZGVyczogIDIuNi4xMS1yNQpBQ0NFUFRfS0VZV09SRFM9Ing4NiIKQVVUT0NMRUFO PSJ5ZXMiCkNCVUlMRD0iaTY4Ni1wYy1saW51eC1nbnUiCkNGTEFHUz0iLW1hcmNoPWF0aGxvbi14 cCAtbWZwbWF0aD1zc2UgLW1zc2UgLU8zIC1waXBlIgpDSE9TVD0iaTY4Ni1wYy1saW51eC1nbnUi CkNPTkZJR19QUk9URUNUPSIvZXRjIC91c3Iva2RlLzMuNS9lbnYgL3Vzci9rZGUvMy41L3NoYXJl L2NvbmZpZyAvdXNyL2tkZS8zLjUvc2h1dGRvd24gL3Vzci9zaGFyZS9YMTEveGtiIC91c3Ivc2hh cmUvY29uZmlnIC91c3Ivc2hhcmUvdGV4bWYvZHZpcGRmbS9jb25maWcvIC91c3Ivc2hhcmUvdGV4 bWYvZHZpcHMvY29uZmlnLyAvdXNyL3NoYXJlL3RleG1mL3RleC9nZW5lcmljL2NvbmZpZy8gL3Vz ci9zaGFyZS90ZXhtZi90ZXgvcGxhdGV4L2NvbmZpZy8gL3Vzci9zaGFyZS90ZXhtZi94ZHZpLyIK Q09ORklHX1BST1RFQ1RfTUFTSz0iL2V0Yy9lbnYuZCAvZXRjL2djb25mIC9ldGMvcmV2ZGVwLXJl YnVpbGQgL2V0Yy90ZXJtaW5mbyIKQ1hYRkxBR1M9Ii1tYXJjaD1hdGhsb24teHAgLW1mcG1hdGg9 c3NlIC1tc3NlIC1PMyAtcGlwZSIKRElTVERJUj0iL3Vzci9wb3J0YWdlL2Rpc3RmaWxlcyIKRkVB VFVSRVM9ImF1dG9jb25maWcgZGlzdGxvY2tzIG1ldGFkYXRhLXRyYW5zZmVyIHNhbmRib3ggc2Zw ZXJtcyBzdHJpY3QiCkdFTlRPT19NSVJST1JTPSJmdHA6Ly9mdHBsLnR1LWNoZW1uaXR6LmRlL3B1 Yi9saW51eC9nZW50b28gICAgICBodHRwOi8vZnRwbC50dS1jaGVtbml0ei5kZS9wdWIvbGludXgv Z2VudG9vICAgICBodHRwOi8vZGlzdGZpbGVzLmdlbnRvby5vcmcgICAgIGh0dHA6Ly93d3cuaWJp Ymxpby5vcmcvcHViL0xpbnV4L2Rpc3RyaWJ1dGlvbnMvZ2VudG9vICAiCkxBTkc9ImRlX0RFLlVU Ri04IgpMQ19BTEw9ImRlX0RFLlVURi04IgpMSU5HVUFTPSJlbiBkZSIKTUFLRU9QVFM9Ii1qMiIK UEtHRElSPSIvdXNyL3BvcnRhZ2UvcGFja2FnZXMiClBPUlRBR0VfUlNZTkNfT1BUUz0iLS1yZWN1 cnNpdmUgLS1saW5rcyAtLXNhZmUtbGlua3MgLS1wZXJtcyAtLXRpbWVzIC0tY29tcHJlc3MgLS1m b3JjZSAtLXdob2xlLWZpbGUgLS1kZWxldGUgLS1kZWxldGUtYWZ0ZXIgLS1zdGF0cyAtLXRpbWVv dXQ9MTgwIC0tZXhjbHVkZT0nL2Rpc3RmaWxlcycgLS1leGNsdWRlPScvbG9jYWwnIC0tZXhjbHVk ZT0nL3BhY2thZ2VzJyIKUE9SVEFHRV9UTVBESVI9Ii92YXIvdG1wIgpQT1JURElSPSIvdXNyL3Bv cnRhZ2UiClNZTkM9InJzeW5jOi8vcnN5bmMuZGUuZ2VudG9vLm9yZy9nZW50b28tcG9ydGFnZSIK VVNFPSJ4ODYgM2Rub3cgM2Rub3dleHQgWCBhYWxpYiBhbHNhIGFwYWNoZTIgYXBtIGFydHMgYXZp IGJhc2gtY29tcGxldGlvbiBiZXJrZGIgYml0bWFwLWZvbnRzIGNsaSBjcnlwdCBjdXBzIGN2cyBk bGxvYWRlciBkcmkgZHZkIGR2ZHJlYWQgZWRzIGVsaWJjX2dsaWJjIGVtYm9zcyBlbmNvZGUgZm9v bWF0aWNkYiBmb3J0cmFuIGdkYm0gZ2lmIGdwbSBnc3RyZWFtZXIgZ3RrIGd0azIgaW1saWIgaW5w dXRfZGV2aWNlc19ldmRldiBpbnB1dF9kZXZpY2VzX2tleWJvYXJkIGlucHV0X2RldmljZXNfbW91 c2UgaXB2NiBpc2RubG9nIGpwZWcga2RlIGtlcm5lbF9saW51eCBsaWJnKysgbGlid3d3IGxpbmd1 YXNfZGUgbGluZ3Vhc19lbiBtYWQgbWlrbW9kIG1teCBtbXhleHQgbW90aWYgbXAzIG1wZWcgbXBs YXllciBuY3Vyc2VzIG5scyBucHRsIG5wdGxvbmx5IG5zcGx1Z2luIG9nZyBvZ2d2b3JiaXMgb3Bl bmdsIG9zcyBwYW0gcGNyZSBwZGZsaWIgcGVybCBwbmcgcHBwZCBweXRob24gcXQzIHF0NCBxdWlj a3RpbWUgcmVhZGxpbmUgcmVhbCByZWZsZWN0aW9uIHNkbCBzZXNzaW9uIHNwZWxsIHNwbCBzc2Ug c3NsIHN1YnZlcnNpb24gdGNwZCB0ZXRleCB0aHJlYWRzIHRydWV0eXBlIHRydWV0eXBlLWZvbnRz IHR5cGUxLWZvbnRzIHVuaWNvZGUgdXNlcmxhbmRfR05VIHZpZGVvX2NhcmRzX252aWRpYSB2b3Ji aXMgeG1sMiB4bW1zIHhvcmcgeHYgemxpYiIKVW5zZXQ6ICBDVEFSR0VULCBFTUVSR0VfREVGQVVM VF9PUFRTLCBJTlNUQUxMX01BU0ssIExERkxBR1MsIFBPUlRBR0VfUlNZTkNfRVhUUkFfT1BUUywg UE9SVERJUl9PVkVSTEFZCg==