| Message ID | 4C4D13B3-179F-44B9-BDB0-29EE706C589A@sandoe-acoustics.co.uk |
|---|---|
| State | New |
| Headers | show |
On Fri, Jun 17, 2011 at 12:04:34PM +0100, IainS wrote: > Hi Jack, > > On 17 Jun 2011, at 03:21, Jack Howarth wrote: > >> >> The gcj compiler needs to pass -no_pie for linkage on darwin11 due to >> the new -pie >> default of the linker. The attached patch accomplishes this by passing >> -no_pie on SYSTEMSPEC >> for *-*-darwin[12]*. Since Darwin10 supports -no_pie in its linker, I >> included it in the >> triplet match to simplify the syntax. Bootstrap and tested on x86_64- >> apple-darwin11. >> Okay for gcc trunk? >> Jack >> >> 2011-06-16 Jack Howarth <howarth@bromo.med.uc.edu> >> >> * libjava/configure.ac (SYSTEMSPEC): Pass -no_pie for darwin11. >> * libjava/configure: Regenerate. > > > I would like to see some more analysis of what the underlying reasons > for failure are. Iain, I suspect these failures are related to those seen in the boehm-gc testsuite when those tests are built with the default -pie linker option. I'll have to repeat a build without the patch but a cursory examination suggests ecj1 is miscompiled with -pie. If I remove -no_pie from /sw/lib/gcc4.7/lib/libgcj.spec after building and installing gcc trunk, the crashes in gcj don't occur. IMHO, my patch is effectively doing the same thing that Mike checked in with... # Ensure we don't try and use -pie, as it is incompatible with pch. BOOT_LDFLAGS += `case ${host} in *-*-darwin[1][1-9]*) echo -Wl,-no_pie ;; esac;` The libjava build uses gcj.spec to build ecj1, etc and Mike's change is insufficient to cause -no_pie to be used for those cases (since those parts of the libjava bootstrap don't honor BOOT_LDFLAGS). Jack ps I'll repeat the bootstrap without my patch and open a PR for the gcj failure. However, I believe it was an exception failure which would place us in the system unwinder and that will be extremely difficult to debug outside of Apple. > > -fpie works fine with darwin 9 and darwin 10 libjava [XCode 3.1.4 and > 3.2.5 respectively, bootstrap w/4.2.1] > (modulo suppressing it when building test-suite .dylibs *** - which is a > testsuite options handling issue - not a fundamental problem). > > so: > > make -k check-target-libjava "RUNTESTFLAGS=--target_board=unix/-fpie\{- > m32,-m64\} " > > passes without regression (re the no pie case) given that pie is > suppressed for dylibs. > > ----- > > So is your proposed patch a work-around for (as yet unreleased) darwin > 11 tool-chain bugs or ... ? > > Iain > > === > **** > kludge to suppress pie for for dylibs (use in place of the darwin9.h > hunk from the attachment on PR49371). > > Index: gcc/config/darwin9.h > =================================================================== > --- gcc/config/darwin9.h (revision 175110) > +++ gcc/config/darwin9.h (working copy) > @@ -35,6 +35,12 @@ along with GCC; see the file COPYING3. If not see > /* Tell collect2 to run dsymutil for us as necessary. */ > #define COLLECT_RUN_DSYMUTIL 1 > > +#undef PIE_SPEC > +#define PIE_SPEC \ > + "%{fpie|pie|fPIE: %{!Zdynamiclib: \ > + %{mdynamic-no-pic: %n'-mdynamic-no-pic' overrides '-pie', '-fpie' > or '-fPIE'; \ > + :-pie}}}" > + > #undef ASM_OUTPUT_ALIGNED_COMMON > #define ASM_OUTPUT_ALIGNED_COMMON(FILE, NAME, SIZE, ALIGN) \ > do { \
On Fri, Jun 17, 2011 at 12:04:34PM +0100, IainS wrote: > Hi Jack, > > On 17 Jun 2011, at 03:21, Jack Howarth wrote: > >> >> The gcj compiler needs to pass -no_pie for linkage on darwin11 due to >> the new -pie >> default of the linker. The attached patch accomplishes this by passing >> -no_pie on SYSTEMSPEC >> for *-*-darwin[12]*. Since Darwin10 supports -no_pie in its linker, I >> included it in the >> triplet match to simplify the syntax. Bootstrap and tested on x86_64- >> apple-darwin11. >> Okay for gcc trunk? >> Jack >> >> 2011-06-16 Jack Howarth <howarth@bromo.med.uc.edu> >> >> * libjava/configure.ac (SYSTEMSPEC): Pass -no_pie for darwin11. >> * libjava/configure: Regenerate. > > > I would like to see some more analysis of what the underlying reasons > for failure are. Iain, Stock gcc trunk without my patch, which creates a gcj (ecj1) linked with -pie, crashes as... [MacPro:~] howarth% gcj-fsf-4.7 --main=testme -O testme.java -pie -v Using built-in specs. Reading specs from /sw/lib/gcc4.7/lib/gcc/x86_64-apple-darwin11.0.0/4.7.0/../../../libgcj.spec rename spec startfile to startfileorig rename spec lib to liborig COLLECT_GCC=gcj-fsf-4.7 COLLECT_LTO_WRAPPER=/sw/lib/gcc4.7/libexec/gcc/x86_64-apple-darwin11.0.0/4.7.0/lto-wrapper Target: x86_64-apple-darwin11.0.0 Configured with: ../gcc-4.7-20110617/configure --prefix=/sw --prefix=/sw/lib/gcc4.7 --mandir=/sw/share/man --infodir=/sw/lib/gcc4.7/info --enable-languages=c,c++,fortran,objc,obj-c++,java --with-gmp=/sw --with-libiconv-prefix=/sw --with-ppl=/sw --with-cloog=/sw --with-mpc=/sw --with-system-zlib --x-includes=/usr/X11R6/include --x-libraries=/usr/X11R6/lib --program-suffix=-fsf-4.7 --enable-checking=yes --enable-cloog-backend=isl Thread model: posix gcc version 4.7.0 20110617 (experimental) (GCC) COLLECT_GCC_OPTIONS='-fsaw-java-file' '-mmacosx-version-min=10.7.0' '-O' '-pie' '-v' '-fbootclasspath=./:/sw/lib/gcc4.7/share/java/libgcj-4.7.0.jar' '-shared-libgcc' '-mtune=core2' /sw/lib/gcc4.7/libexec/gcc/x86_64-apple-darwin11.0.0/4.7.0/ecj1 testme.java -fbootclasspath=./:/sw/lib/gcc4.7/share/java/libgcj-4.7.0.jar -fsource=1.5 -ftarget=1.5 -fzip-dependency /var/folders/1l/n78sywl52lz6kkys6nv7mnph0000gp/T//ccxj1zOQ.zip -fzip-target /var/folders/1l/n78sywl52lz6kkys6nv7mnph0000gp/T//ccXXudTY.jar Exception in thread "main" java.lang.ClassFormatError: org.eclipse.jdt.internal.compiler.Compiler (erroneous exception handler info) at java.lang.VMClassLoader.defineClass(libgcj.12.dylib) at java.lang.ClassLoader.defineClass(libgcj.12.dylib) at java.security.SecureClassLoader.defineClass(libgcj.12.dylib) at java.net.URLClassLoader.findClass(libgcj.12.dylib) at java.lang.ClassLoader.loadClass(libgcj.12.dylib) at java.lang.ClassLoader.loadClass(libgcj.12.dylib) at org.eclipse.jdt.internal.compiler.impl.CompilerOptions.resetDefaults(CompilerOptions.java:963) at org.eclipse.jdt.internal.compiler.impl.CompilerOptions.<init>(CompilerOptions.java:371) at org.eclipse.jdt.internal.compiler.impl.CompilerOptions.<init>(CompilerOptions.java:363) at org.eclipse.jdt.internal.compiler.batch.Main.initialize(Main.java:3548) at org.eclipse.jdt.internal.compiler.batch.Main.<init>(Main.java:1435) at org.eclipse.jdt.internal.compiler.batch.Main.<init>(Main.java:1423) at org.eclipse.jdt.internal.compiler.batch.GCCMain.<init>(GCCMain.java:62) at org.eclipse.jdt.internal.compiler.batch.GCCMain.main(GCCMain.java:498) Running... MacPro:~] howarth% gdb /sw/lib/gcc4.7/libexec/gcc/x86_64-apple-darwin11.0.0/4.7.0/ecj1 ... (gdb) r testme.java -fbootclasspath=./:/sw/lib/gcc4.7/share/java/libgcj-4.7.0.jar -fsource=1.5 -ftarget=1.5 -fzip-dependency testme.zip -fzip-target testme.jar Starting program: /sw/lib/gcc4.7/libexec/gcc/x86_64-apple-darwin11.0.0/4.7.0/ecj1 testme.java -fbootclasspath=./:/sw/lib/gcc4.7/share/java/libgcj-4.7.0.jar -fsource=1.5 -ftarget=1.5 -fzip-dependency testme.zip -fzip-target testme.jar Reading symbols for shared libraries +++++........................ done Program exited normally. ...produces no backtraces (and no a.out). Jack > > -fpie works fine with darwin 9 and darwin 10 libjava [XCode 3.1.4 and > 3.2.5 respectively, bootstrap w/4.2.1] > (modulo suppressing it when building test-suite .dylibs *** - which is a > testsuite options handling issue - not a fundamental problem). > > so: > > make -k check-target-libjava "RUNTESTFLAGS=--target_board=unix/-fpie\{- > m32,-m64\} " > > passes without regression (re the no pie case) given that pie is > suppressed for dylibs. > > ----- > > So is your proposed patch a work-around for (as yet unreleased) darwin > 11 tool-chain bugs or ... ? > > Iain > > === > **** > kludge to suppress pie for for dylibs (use in place of the darwin9.h > hunk from the attachment on PR49371). > > Index: gcc/config/darwin9.h > =================================================================== > --- gcc/config/darwin9.h (revision 175110) > +++ gcc/config/darwin9.h (working copy) > @@ -35,6 +35,12 @@ along with GCC; see the file COPYING3. If not see > /* Tell collect2 to run dsymutil for us as necessary. */ > #define COLLECT_RUN_DSYMUTIL 1 > > +#undef PIE_SPEC > +#define PIE_SPEC \ > + "%{fpie|pie|fPIE: %{!Zdynamiclib: \ > + %{mdynamic-no-pic: %n'-mdynamic-no-pic' overrides '-pie', '-fpie' > or '-fPIE'; \ > + :-pie}}}" > + > #undef ASM_OUTPUT_ALIGNED_COMMON > #define ASM_OUTPUT_ALIGNED_COMMON(FILE, NAME, SIZE, ALIGN) \ > do { \
On Fri, Jun 17, 2011 at 12:04:34PM +0100, IainS wrote: > Hi Jack, > > On 17 Jun 2011, at 03:21, Jack Howarth wrote: > >> >> The gcj compiler needs to pass -no_pie for linkage on darwin11 due to >> the new -pie >> default of the linker. The attached patch accomplishes this by passing >> -no_pie on SYSTEMSPEC >> for *-*-darwin[12]*. Since Darwin10 supports -no_pie in its linker, I >> included it in the >> triplet match to simplify the syntax. Bootstrap and tested on x86_64- >> apple-darwin11. >> Okay for gcc trunk? >> Jack >> >> 2011-06-16 Jack Howarth <howarth@bromo.med.uc.edu> >> >> * libjava/configure.ac (SYSTEMSPEC): Pass -no_pie for darwin11. >> * libjava/configure: Regenerate. > > > I would like to see some more analysis of what the underlying reasons > for failure are. > > -fpie works fine with darwin 9 and darwin 10 libjava [XCode 3.1.4 and > 3.2.5 respectively, bootstrap w/4.2.1] > (modulo suppressing it when building test-suite .dylibs *** - which is a > testsuite options handling issue - not a fundamental problem). This is from a complete bootstrap with... BOOT_LDFLAGS += -Wl,-no_pie Otherwise you are not duplicating darwin11's -pie linker default. Again it is essential that you cause ecj1 to be linked with -pie to manifest this problem. Also note the comments in... http://gcc.gnu.org/ml/gcc/2011-06/msg00211.html which argue that there is little to be gained by building gcc as PIE. If that is valid, my patch achieves that. The problem is that the infrastructure for gcj and ecj1 currently doesn't allow BOOT_LDFLAGS to be used when ecj1 is built. Note that if I go back to the build directory (without my patch), relink ecjx with -Wl,-no_pie and copy that over the installed ecj1, the crash in gcj disappears. Jack > > so: > > make -k check-target-libjava "RUNTESTFLAGS=--target_board=unix/-fpie\{- > m32,-m64\} " > > passes without regression (re the no pie case) given that pie is > suppressed for dylibs. > > ----- > > So is your proposed patch a work-around for (as yet unreleased) darwin > 11 tool-chain bugs or ... ? > > Iain > > === > **** > kludge to suppress pie for for dylibs (use in place of the darwin9.h > hunk from the attachment on PR49371). > > Index: gcc/config/darwin9.h > =================================================================== > --- gcc/config/darwin9.h (revision 175110) > +++ gcc/config/darwin9.h (working copy) > @@ -35,6 +35,12 @@ along with GCC; see the file COPYING3. If not see > /* Tell collect2 to run dsymutil for us as necessary. */ > #define COLLECT_RUN_DSYMUTIL 1 > > +#undef PIE_SPEC > +#define PIE_SPEC \ > + "%{fpie|pie|fPIE: %{!Zdynamiclib: \ > + %{mdynamic-no-pic: %n'-mdynamic-no-pic' overrides '-pie', '-fpie' > or '-fPIE'; \ > + :-pie}}}" > + > #undef ASM_OUTPUT_ALIGNED_COMMON > #define ASM_OUTPUT_ALIGNED_COMMON(FILE, NAME, SIZE, ALIGN) \ > do { \
On Fri, Jun 17, 2011 at 12:04:34PM +0100, IainS wrote: > Hi Jack, > > On 17 Jun 2011, at 03:21, Jack Howarth wrote: > >> >> The gcj compiler needs to pass -no_pie for linkage on darwin11 due to >> the new -pie >> default of the linker. The attached patch accomplishes this by passing >> -no_pie on SYSTEMSPEC >> for *-*-darwin[12]*. Since Darwin10 supports -no_pie in its linker, I >> included it in the >> triplet match to simplify the syntax. Bootstrap and tested on x86_64- >> apple-darwin11. >> Okay for gcc trunk? >> Jack >> >> 2011-06-16 Jack Howarth <howarth@bromo.med.uc.edu> >> >> * libjava/configure.ac (SYSTEMSPEC): Pass -no_pie for darwin11. >> * libjava/configure: Regenerate. > > > I would like to see some more analysis of what the underlying reasons > for failure are. > > -fpie works fine with darwin 9 and darwin 10 libjava [XCode 3.1.4 and > 3.2.5 respectively, bootstrap w/4.2.1] > (modulo suppressing it when building test-suite .dylibs *** - which is a > testsuite options handling issue - not a fundamental problem). > > so: > > make -k check-target-libjava "RUNTESTFLAGS=--target_board=unix/-fpie\{- > m32,-m64\} " > > passes without regression (re the no pie case) given that pie is > suppressed for dylibs. > > ----- > > So is your proposed patch a work-around for (as yet unreleased) darwin > 11 tool-chain bugs or ... ? > > Iain > > === > **** > kludge to suppress pie for for dylibs (use in place of the darwin9.h > hunk from the attachment on PR49371). Iain, This change should be unncessary since -pie on the linker only operates on executables (hence the name PIE). Jack > > Index: gcc/config/darwin9.h > =================================================================== > --- gcc/config/darwin9.h (revision 175110) > +++ gcc/config/darwin9.h (working copy) > @@ -35,6 +35,12 @@ along with GCC; see the file COPYING3. If not see > /* Tell collect2 to run dsymutil for us as necessary. */ > #define COLLECT_RUN_DSYMUTIL 1 > > +#undef PIE_SPEC > +#define PIE_SPEC \ > + "%{fpie|pie|fPIE: %{!Zdynamiclib: \ > + %{mdynamic-no-pic: %n'-mdynamic-no-pic' overrides '-pie', '-fpie' > or '-fPIE'; \ > + :-pie}}}" > + > #undef ASM_OUTPUT_ALIGNED_COMMON > #define ASM_OUTPUT_ALIGNED_COMMON(FILE, NAME, SIZE, ALIGN) \ > do { \
Hi Jack, On 18 Jun 2011, at 03:57, Jack Howarth wrote: > On Fri, Jun 17, 2011 at 12:04:34PM +0100, IainS wrote: >> Hi Jack, >> === >> **** >> kludge to suppress pie for for dylibs (use in place of the darwin9.h >> hunk from the attachment on PR49371). > > Iain, > This change should be unncessary since -pie on the linker only > operates > on executables (hence the name PIE). On darwin < 11 .. which don't default to linking pie: As stated, it's a kludge to allow fpie to be put as a test-suite option... ... since the options are passed verbatim to library builds, all the dylib builds fail otherwise ... if there is an alternate way of filtering 'fpie/fPIE' out of dylib builds, that could be more globally applied the to the test suite, that would be a good fix to make. (the kludge is expedient to test the hypothesis that pie works with libjava on darwin 9 & 10 -- FAOD: I am not suggesting it for trunk)... Iain >> Index: gcc/config/darwin9.h >> =================================================================== >> --- gcc/config/darwin9.h (revision 175110) >> +++ gcc/config/darwin9.h (working copy) >> @@ -35,6 +35,12 @@ along with GCC; see the file COPYING3. If not see >> /* Tell collect2 to run dsymutil for us as necessary. */ >> #define COLLECT_RUN_DSYMUTIL 1 >> >> +#undef PIE_SPEC >> +#define PIE_SPEC \ >> + "%{fpie|pie|fPIE: %{!Zdynamiclib: \ >> + %{mdynamic-no-pic: %n'-mdynamic-no-pic' overrides '-pie', '- >> fpie' >> or '-fPIE'; \ >> + :-pie}}}" >> + >> #undef ASM_OUTPUT_ALIGNED_COMMON >> #define ASM_OUTPUT_ALIGNED_COMMON(FILE, NAME, SIZE, >> ALIGN) \ >> do >> { \
=== **** kludge to suppress pie for for dylibs (use in place of the darwin9.h hunk from the attachment on PR49371). Index: gcc/config/darwin9.h =================================================================== --- gcc/config/darwin9.h (revision 175110) +++ gcc/config/darwin9.h (working copy) @@ -35,6 +35,12 @@ along with GCC; see the file COPYING3. If not see /* Tell collect2 to run dsymutil for us as necessary. */ #define COLLECT_RUN_DSYMUTIL 1 +#undef PIE_SPEC +#define PIE_SPEC \ + "%{fpie|pie|fPIE: %{!Zdynamiclib: \ + %{mdynamic-no-pic: %n'-mdynamic-no-pic' overrides '-pie', '- fpie' or '-fPIE'; \ + :-pie}}}" + #undef ASM_OUTPUT_ALIGNED_COMMON #define ASM_OUTPUT_ALIGNED_COMMON(FILE, NAME, SIZE,