156015 2016-03-30 01:33:26 -0700 Fails to build in Linux / PowerPC due to different ucontext_t definition 2016-03-31 14:52:08 -0700 1 1 1 Unclassified WebKit JavaScriptCore Other Unspecified Unspecified RESOLVED FIXED P2 Normal --- 1 berto ysuzuki berto cgarcia commit-queue keith_miller mark.lam mcatanzaro msaboff saam ysuzuki oldest_to_newest 1179019 0 berto 2016-03-30 01:33:26 -0700 /«PKGBUILDDIR»/Source/JavaScriptCore/heap/MachineStackMarker.cpp: In function 'void pthreadSignalHandlerSuspendResume(int, siginfo_t*, void*)': /«PKGBUILDDIR»/Source/JavaScriptCore/heap/MachineStackMarker.cpp:89:37: error: no match for 'operator=' (operand types are 'mcontext_t' and 'ucontext::uc_regs_ptr') thread->suspendedMachineContext = userContext->uc_mcontext; ^ In file included from /usr/include/signal.h:326:0, from /«PKGBUILDDIR»/Source/JavaScriptCore/heap/MachineStackMarker.h:36, from /«PKGBUILDDIR»/Source/JavaScriptCore/heap/MachineStackMarker.cpp:23: /usr/include/powerpc-linux-gnu/sys/ucontext.h:60:3: note: candidate: mcontext_t& mcontext_t::operator=(const mcontext_t&) } mcontext_t; ^ /usr/include/powerpc-linux-gnu/sys/ucontext.h:60:3: note: no known conversion for argument 1 from 'ucontext::uc_regs_ptr' to 'const mcontext_t&' /usr/include/powerpc-linux-gnu/sys/ucontext.h:60:3: note: candidate: mcontext_t& mcontext_t::operator=(mcontext_t&&) /usr/include/powerpc-linux-gnu/sys/ucontext.h:60:3: note: no known conversion for argument 1 from 'ucontext::uc_regs_ptr' to 'mcontext_t&&' Source/JavaScriptCore/CMakeFiles/JavaScriptCore.dir/build.make:9801: recipe for target 'Source/JavaScriptCore/CMakeFiles/JavaScriptCore.dir/heap/MachineStackMarker.cpp.o' failed make[3]: *** [Source/JavaScriptCore/CMakeFiles/JavaScriptCore.dir/heap/MachineStackMarker.cpp.o] Error 1 s 1179020 1 berto 2016-03-30 01:36:44 -0700 [ somehow the previous comment is incomplete, please ignore it ] I reproduced this with WebKitGTK+ 2.12.0. It seems that the definition of ucontext_t is different in Linux PowerPC, producing this build error: /«PKGBUILDDIR»/Source/JavaScriptCore/heap/MachineStackMarker.cpp: In function 'void pthreadSignalHandlerSuspendResume(int, siginfo_t*, void*)': /«PKGBUILDDIR»/Source/JavaScriptCore/heap/MachineStackMarker.cpp:89:37: error: no match for 'operator=' (operand types are 'mcontext_t' and 'ucontext::uc_regs_ptr') thread->suspendedMachineContext = userContext->uc_mcontext; ^ In file included from /usr/include/signal.h:326:0, from /«PKGBUILDDIR»/Source/JavaScriptCore/heap/MachineStackMarker.h:36, from /«PKGBUILDDIR»/Source/JavaScriptCore/heap/MachineStackMarker.cpp:23: /usr/include/powerpc-linux-gnu/sys/ucontext.h:60:3: note: candidate: mcontext_t& mcontext_t::operator=(const mcontext_t&) } mcontext_t; ^ /usr/include/powerpc-linux-gnu/sys/ucontext.h:60:3: note: no known conversion for argument 1 from 'ucontext::uc_regs_ptr' to 'const mcontext_t&' /usr/include/powerpc-linux-gnu/sys/ucontext.h:60:3: note: candidate: mcontext_t& mcontext_t::operator=(mcontext_t&&) /usr/include/powerpc-linux-gnu/sys/ucontext.h:60:3: note: no known conversion for argument 1 from 'ucontext::uc_regs_ptr' to 'mcontext_t&&' Source/JavaScriptCore/CMakeFiles/JavaScriptCore.dir/build.make:9801: recipe for target 'Source/JavaScriptCore/CMakeFiles/JavaScriptCore.dir/heap/MachineStackMarker.cpp.o' failed make[3]: *** [Source/JavaScriptCore/CMakeFiles/JavaScriptCore.dir/heap/MachineStackMarker.cpp.o] Error 1 Here's how it is defined: typedef struct ucontext { struct ucontext *uc_link; sigset_t uc_sigmask; stack_t uc_stack; mcontext_t uc_mcontext; ... } ucontext_t; Here's how it is defined in powerpc: typedef struct ucontext { struct ucontext *uc_link; sigset_t uc_sigmask; stack_t uc_stack; union uc_regs_ptr { struct pt_regs *regs; mcontext_t *uc_regs; } uc_mcontext; ... } ucontext_t; Full definition here: https://sources.debian.net/src/glibc/2.21-9/sysdeps/unix/sysv/linux/powerpc/sys/ucontext.h/?hl=139#L139 According to the setcontext(3) man page, this was removed in POSIX.1-2008 because of portability reasons. 1179384 2 275275 ysuzuki 2016-03-31 06:09:39 -0700 Created attachment 275275 Patch 1179385 3 ysuzuki 2016-03-31 06:10:18 -0700 Could you ensure that the uploaded patch fixes the issue? I don't have any PowerPC machine. 1179412 4 berto 2016-03-31 08:29:51 -0700 (In reply to comment #3) > Could you ensure that the uploaded patch fixes the issue? I don't have any > PowerPC machine. Hey, thanks for the patch! I'm currently building it and will test it later. I noticed however that with this patch what suspendedMachineContext stores is not an mcontext_t but a pointer to an mcontext_t. If I understand it correctly the actual mcontext_t data in the PowerPC case is stored in ucontext_t.uc_reg_space. I wonder if this patch is correct then? If not maybe there's no portable way to do this... 1179416 5 ysuzuki 2016-03-31 08:37:27 -0700 (In reply to comment #4) > (In reply to comment #3) > > Could you ensure that the uploaded patch fixes the issue? I don't have any > > PowerPC machine. > > Hey, thanks for the patch! > > I'm currently building it and will test it later. I noticed however > that with this patch what suspendedMachineContext stores is not an > mcontext_t but a pointer to an mcontext_t. If I understand it > correctly the actual mcontext_t data in the PowerPC case is stored in > ucontext_t.uc_reg_space. > > I wonder if this patch is correct then? If not maybe there's no > portable way to do this... Oops, that's right. I'll change the patch soon. 1179419 6 275295 ysuzuki 2016-03-31 08:44:52 -0700 Created attachment 275295 Patch 1179420 7 ysuzuki 2016-03-31 08:45:13 -0700 (In reply to comment #6) > Created attachment 275295 [details] > Patch OK, could you try it again? 1179464 8 berto 2016-03-31 10:42:04 -0700 (In reply to comment #7) > (In reply to comment #6) > > Created attachment 275295 [details] > > Patch > > OK, could you try it again? Yeah, that works. 1179517 9 275295 mcatanzaro 2016-03-31 11:58:44 -0700 Comment on attachment 275295 Patch View in context: https://bugs.webkit.org/attachment.cgi?id=275295&action=review > Source/JavaScriptCore/heap/MachineStackMarker.cpp:90 > + // Since glib in PowerPC does not have mcontext_t in ucontext_t::uc_mcontext, we specially handle this case here. I would omit this comment; the conditional compilation makes it obvious that the data structure is architecture-specific. 1179595 10 275295 ysuzuki 2016-03-31 14:48:50 -0700 Comment on attachment 275295 Patch View in context: https://bugs.webkit.org/attachment.cgi?id=275295&action=review >> Source/JavaScriptCore/heap/MachineStackMarker.cpp:90 >> + // Since glib in PowerPC does not have mcontext_t in ucontext_t::uc_mcontext, we specially handle this case here. > > I would omit this comment; the conditional compilation makes it obvious that the data structure is architecture-specific. OK, thanks. Omitted. 1179599 11 ysuzuki 2016-03-31 14:52:08 -0700 Committed r198919: <http://trac.webkit.org/changeset/198919> 275275 2016-03-31 06:09:39 -0700 2016-03-31 08:44:48 -0700 Patch bug-156015-20160331220942.patch text/plain 2218 ysuzuki U3VidmVyc2lvbiBSZXZpc2lvbjogMTk4ODkxCmRpZmYgLS1naXQgYS9Tb3VyY2UvSmF2YVNjcmlw dENvcmUvQ2hhbmdlTG9nIGIvU291cmNlL0phdmFTY3JpcHRDb3JlL0NoYW5nZUxvZwppbmRleCBk YzE5ZTg2ZDZmNzE4MDM3NjU2Y2M4NjcxNmVlZTY1ZjEwM2E1OTcwLi4wMWIyYjY1YjM4NDIwNjM5 ZDQyYzdmMjkzZjczODQxZGRkZDc3MDlkIDEwMDY0NAotLS0gYS9Tb3VyY2UvSmF2YVNjcmlwdENv cmUvQ2hhbmdlTG9nCisrKyBiL1NvdXJjZS9KYXZhU2NyaXB0Q29yZS9DaGFuZ2VMb2cKQEAgLTEs MyArMSwxNiBAQAorMjAxNi0wMy0zMSAgWXVzdWtlIFN1enVraSAgPHV0YXRhbmUudGVhQGdtYWls LmNvbT4KKworICAgICAgICBGYWlscyB0byBidWlsZCBpbiBMaW51eCAvIFBvd2VyUEMgZHVlIHRv IGRpZmZlcmVudCB1Y29udGV4dF90IGRlZmluaXRpb24KKyAgICAgICAgaHR0cHM6Ly9idWdzLndl YmtpdC5vcmcvc2hvd19idWcuY2dpP2lkPTE1NjAxNQorCisgICAgICAgIFJldmlld2VkIGJ5IE5P Qk9EWSAoT09QUyEpLgorCisgICAgICAgIFBQQyBkb2VzIG5vdCBoYXZlIG1jb250ZXh0X3QgaW4g dWNvbnRleHRfdDo6dWNfbWNvbnRleHQuCisgICAgICAgIFNvIGluc3RlYWQgb2YgdXNpbmcgbWNv bnRleHRfdCBkaXJlY3RseSwgd2UgcmV0cmlldmUgdGhlIHR5cGUgb2YgdWNvbnRleHRfdDo6dWNf bWNvbnRleHQgYW5kCisgICAgICAgIG5hbWUgdGhhdCB0eXBlIGFzICJNYWNoaW5lQ29udGV4dCIu CisKKyAgICAgICAgKiBoZWFwL01hY2hpbmVTdGFja01hcmtlci5oOgorCiAyMDE2LTAzLTMwICBC ZW5qYW1pbiBQb3VsYWluICA8YmVuamFtaW5Ad2Via2l0Lm9yZz4KIAogICAgICAgICBbSlNDXVt4 ODZdIEFkZCB0aGUgMyBvcGVyYW5kcyBmb3JtcyBvZiBmbG9hdGluZyBwb2ludCBhZGRpdGlvbiBh bmQgbXVsdGlwbGljYXRpb24KZGlmZiAtLWdpdCBhL1NvdXJjZS9KYXZhU2NyaXB0Q29yZS9oZWFw L01hY2hpbmVTdGFja01hcmtlci5oIGIvU291cmNlL0phdmFTY3JpcHRDb3JlL2hlYXAvTWFjaGlu ZVN0YWNrTWFya2VyLmgKaW5kZXggYTE2ZjBkYWUxNjBlMjJkMTlkNmQ0NzU5MjQ1NDM0MDA0ZTky NGY4MS4uZGNiOGZlYzA5MmVmMGZkYWE2YjgwNWQ4MTJkNzA5OGQyYmRkMDc2MiAxMDA2NDQKLS0t IGEvU291cmNlL0phdmFTY3JpcHRDb3JlL2hlYXAvTWFjaGluZVN0YWNrTWFya2VyLmgKKysrIGIv U291cmNlL0phdmFTY3JpcHRDb3JlL2hlYXAvTWFjaGluZVN0YWNrTWFya2VyLmgKQEAgLTEwNyw5 ICsxMDcsMTIgQEAgY2xhc3MgTWFjaGluZVRocmVhZHMgewogI2VsaWYgT1MoV0lORE9XUykKICAg ICAgICAgICAgIHR5cGVkZWYgQ09OVEVYVCBQbGF0Zm9ybVJlZ2lzdGVyczsKICNlbGlmIFVTRShQ VEhSRUFEUykKKyAgICAgICAgICAgIC8vIFNpbmNlIGdsaWIgaW4gUG93ZXJQQyBkb2VzIG5vdCBo YXZlIG1jb250ZXh0X3QsIHdlIHJldHJpZXZlIHRoZSB0eXBlIHVzZWQgaW4KKyAgICAgICAgICAg IC8vIHVjb250ZXh0X3Q6OnVjX21jb250ZXh0IGZpZWxkLgorICAgICAgICAgICAgdHlwZWRlZiBk ZWNsdHlwZShzdGF0aWNfY2FzdDx1Y29udGV4dF90Kj4obnVsbHB0ciktPnVjX21jb250ZXh0KSBN YWNoaW5lQ29udGV4dDsKICAgICAgICAgICAgIHN0cnVjdCBQbGF0Zm9ybVJlZ2lzdGVycyB7CiAg ICAgICAgICAgICAgICAgcHRocmVhZF9hdHRyX3QgYXR0cmlidXRlOwotICAgICAgICAgICAgICAg IG1jb250ZXh0X3QgbWFjaGluZUNvbnRleHQ7CisgICAgICAgICAgICAgICAgTWFjaGluZUNvbnRl eHQgbWFjaGluZUNvbnRleHQ7CiAgICAgICAgICAgICB9OwogI2Vsc2UKICNlcnJvciBOZWVkIGEg dGhyZWFkIHJlZ2lzdGVyIHN0cnVjdCBmb3IgdGhpcyBwbGF0Zm9ybQpAQCAtMTM1LDcgKzEzOCw3 IEBAIGNsYXNzIE1hY2hpbmVUaHJlYWRzIHsKICAgICAgICAgSEFORExFIHBsYXRmb3JtVGhyZWFk SGFuZGxlOwogI2VsaWYgVVNFKFBUSFJFQURTKSAmJiAhT1MoREFSV0lOKQogICAgICAgICBzZW1f dCBzZW1hcGhvcmVGb3JTdXNwZW5kUmVzdW1lOwotICAgICAgICBtY29udGV4dF90IHN1c3BlbmRl ZE1hY2hpbmVDb250ZXh0OworICAgICAgICBSZWdpc3RlcnM6Ok1hY2hpbmVDb250ZXh0IHN1c3Bl bmRlZE1hY2hpbmVDb250ZXh0OwogICAgICAgICBpbnQgc3VzcGVuZENvdW50IHsgMCB9OwogICAg ICAgICBzdGQ6OmF0b21pYzxib29sPiBzdXNwZW5kZWQgeyBmYWxzZSB9OwogI2VuZGlmCg== 275295 2016-03-31 08:44:52 -0700 2016-03-31 11:58:44 -0700 Patch bug-156015-20160401004456.patch text/plain 1853 ysuzuki U3VidmVyc2lvbiBSZXZpc2lvbjogMTk4ODkxCmRpZmYgLS1naXQgYS9Tb3VyY2UvSmF2YVNjcmlw dENvcmUvQ2hhbmdlTG9nIGIvU291cmNlL0phdmFTY3JpcHRDb3JlL0NoYW5nZUxvZwppbmRleCBk YzE5ZTg2ZDZmNzE4MDM3NjU2Y2M4NjcxNmVlZTY1ZjEwM2E1OTcwLi4yNDRiYzJmYTkwNTE2MTZi ZjQ5MWE3OTk3ZGY1MTdiODc0NDY1ODIyIDEwMDY0NAotLS0gYS9Tb3VyY2UvSmF2YVNjcmlwdENv cmUvQ2hhbmdlTG9nCisrKyBiL1NvdXJjZS9KYXZhU2NyaXB0Q29yZS9DaGFuZ2VMb2cKQEAgLTEs MyArMSwxNiBAQAorMjAxNi0wMy0zMSAgWXVzdWtlIFN1enVraSAgPHV0YXRhbmUudGVhQGdtYWls LmNvbT4KKworICAgICAgICBGYWlscyB0byBidWlsZCBpbiBMaW51eCAvIFBvd2VyUEMgZHVlIHRv IGRpZmZlcmVudCB1Y29udGV4dF90IGRlZmluaXRpb24KKyAgICAgICAgaHR0cHM6Ly9idWdzLndl YmtpdC5vcmcvc2hvd19idWcuY2dpP2lkPTE1NjAxNQorCisgICAgICAgIFJldmlld2VkIGJ5IE5P Qk9EWSAoT09QUyEpLgorCisgICAgICAgIFBQQyBkb2VzIG5vdCBoYXZlIG1jb250ZXh0X3QgaW4g dWNvbnRleHRfdDo6dWNfbWNvbnRleHQuCisgICAgICAgIFNvIHdlIHRha2UgdGhlIHNwZWNpYWwg d2F5IHRvIHJldHJpZXZlIG1jb250ZXh0X3QgaW4gUFBDLgorCisgICAgICAgICogaGVhcC9NYWNo aW5lU3RhY2tNYXJrZXIuY3BwOgorICAgICAgICAocHRocmVhZFNpZ25hbEhhbmRsZXJTdXNwZW5k UmVzdW1lKToKKwogMjAxNi0wMy0zMCAgQmVuamFtaW4gUG91bGFpbiAgPGJlbmphbWluQHdlYmtp dC5vcmc+CiAKICAgICAgICAgW0pTQ11beDg2XSBBZGQgdGhlIDMgb3BlcmFuZHMgZm9ybXMgb2Yg ZmxvYXRpbmcgcG9pbnQgYWRkaXRpb24gYW5kIG11bHRpcGxpY2F0aW9uCmRpZmYgLS1naXQgYS9T b3VyY2UvSmF2YVNjcmlwdENvcmUvaGVhcC9NYWNoaW5lU3RhY2tNYXJrZXIuY3BwIGIvU291cmNl L0phdmFTY3JpcHRDb3JlL2hlYXAvTWFjaGluZVN0YWNrTWFya2VyLmNwcAppbmRleCAxZGYyODc0 OWY0MDVmMWE0YWUxZmFmODQ4OTc4OGQwZTk2MDUwODgyLi5jZjgyOTI2MTA0NTFlN2NiMTU5YzA3 ZTM5OWJhMTBiNjMzNmM1NjcyIDEwMDY0NAotLS0gYS9Tb3VyY2UvSmF2YVNjcmlwdENvcmUvaGVh cC9NYWNoaW5lU3RhY2tNYXJrZXIuY3BwCisrKyBiL1NvdXJjZS9KYXZhU2NyaXB0Q29yZS9oZWFw L01hY2hpbmVTdGFja01hcmtlci5jcHAKQEAgLTg2LDcgKzg2LDEyIEBAIHN0YXRpYyB2b2lkIHB0 aHJlYWRTaWduYWxIYW5kbGVyU3VzcGVuZFJlc3VtZShpbnQsIHNpZ2luZm9fdCosIHZvaWQqIHVj b250ZXh0KQogICAgIH0KIAogICAgIHVjb250ZXh0X3QqIHVzZXJDb250ZXh0ID0gc3RhdGljX2Nh c3Q8dWNvbnRleHRfdCo+KHVjb250ZXh0KTsKKyNpZiBDUFUoUFBDKQorICAgIC8vIFNpbmNlIGds aWIgaW4gUG93ZXJQQyBkb2VzIG5vdCBoYXZlIG1jb250ZXh0X3QgaW4gdWNvbnRleHRfdDo6dWNf bWNvbnRleHQsIHdlIHNwZWNpYWxseSBoYW5kbGUgdGhpcyBjYXNlIGhlcmUuCisgICAgdGhyZWFk LT5zdXNwZW5kZWRNYWNoaW5lQ29udGV4dCA9ICp1c2VyQ29udGV4dC0+dWNfbWNvbnRleHQudWNf cmVnczsKKyNlbHNlCiAgICAgdGhyZWFkLT5zdXNwZW5kZWRNYWNoaW5lQ29udGV4dCA9IHVzZXJD b250ZXh0LT51Y19tY29udGV4dDsKKyNlbmRpZgogCiAgICAgLy8gQWxsb3cgc3VzcGVuZCBjYWxs ZXIgdG8gc2VlIHRoYXQgdGhpcyB0aHJlYWQgaXMgc3VzcGVuZGVkLgogICAgIC8vIHNlbV9wb3N0 IGlzIGFzeW5jLXNpZ25hbC1zYWZlIGZ1bmN0aW9uLiBJdCBtZWFucyB0aGF0IHdlIGNhbiBjYWxs IHRoaXMgZnJvbSBhIHNpZ25hbCBoYW5kbGVyLgo=