| Message ID | 1465211659-23861-1-git-send-email-vzakhar@synopsys.com |
|---|---|
| State | Accepted |
| Commit | caf515e25e699eb306d0b24986734790de80af28 |
| Headers | show |
>>>>> "Zakharov" == Zakharov Vlad <Vladislav.Zakharov@synopsys.com> writes: > Build failed when "makeinfo" was missing on the build host. > This was happening because "makeinfo" is required to build .info targets > and make exited with error. The issue hadn't appeared before as there was > prebuilt documentation in ARC binutils tarballs, so no attempts had been > made to build docs. > Missing "makeinfo" only stops us from building docs > ("missing" script already throws a warning on that regard). > Let's continue to build other targets. > Now exit code of the script called "missing" is checked. > The value 127 means that "makeinfo" is not available on the build host. > So when such value occurs, 0 is returned to the top level makefile. > Documentation is not being built but further build of binutils continues. > Fixes http://autobuild.buildroot.net/results/55c/55cd09a559016f4f252f0e4c27313b9806135cf4// > Signed-off-by: Zakharov Vlad <vzakhar@synopsys.com> Committed, thanks.
Hello, On Mon, 06 Jun 2016 21:36:54 +0200, Peter Korsgaard wrote: > > Signed-off-by: Zakharov Vlad <vzakhar@synopsys.com> > > Committed, thanks. Why? This problem also exists for other versions of binutils, and also for gdb. And we have a patch series from Romain Naour sitting in patchwork for several weeks. I don't know if Zlad's version is better or not than Romain's version. But at least Romain's version was handling all binutils and gdb versions, without patching directly Makefile.in files. So at first sight, it looked a lot better than Zlad's version. Thomas
>>>>> "Thomas" == Thomas Petazzoni <thomas.petazzoni@free-electrons.com> writes: > Hello, > On Mon, 06 Jun 2016 21:36:54 +0200, Peter Korsgaard wrote: >> > Signed-off-by: Zakharov Vlad <vzakhar@synopsys.com> >> >> Committed, thanks. > Why? It was fixing autobuilder issues and seemed like a sensible fix that could be upstreamed. Looking again, I do see that it patched Makefile.in and not Makefile.am, so that's not too nice though. > This problem also exists for other versions of binutils, and also for > gdb. And we have a patch series from Romain Naour sitting in patchwork > for several weeks. > I don't know if Zlad's version is better or not than Romain's version. > But at least Romain's version was handling all binutils and gdb > versions, without patching directly Makefile.in files. So at first > sight, it looked a lot better than Zlad's version. Ok, we can always revert if it isn't needed any more once Romain's series is applied.
Hi Peter, Thomas, On Mon, 2016-06-06 at 23:34 +0200, Peter Korsgaard wrote: > > > > > > > > > > > > > > > > > > > > > > > > > > "Thomas" == Thomas Petazzoni <thomas.petazzoni@free-electrons.com> writes: > > Hello, > > On Mon, 06 Jun 2016 21:36:54 +0200, Peter Korsgaard wrote: > > >> > Signed-off-by: Zakharov Vlad <vzakhar@synopsys.com> > >> > >> Committed, thanks. > > > Why? > > It was fixing autobuilder issues and seemed like a sensible fix that > could be upstreamed. Looking again, I do see that it patched Makefile.in > and not Makefile.am, so that's not too nice though. Indeed our goal was to fix an issue that causes all autobuilder jobs for ARC to fail. The problem is host binutils couldn't be built any longer after we switched to arc-2016.03 tools. And now we're unblocked. As for patching Makefile.in vs Makefile.am again that's just to do a minimal fix in existing sources. Otherwise if we patch Makefile.am we'll need to regenerate Makefile.in. Even though Vlad has already sent the same patch to Binutils mailing list (https://sourceware.org/ml/binutils/2016-06/msg00053.html) that's indeed not the right fix - he'll need to fix Makefile.am but at least we're moving in upstreaming direction. > > This problem also exists for other versions of binutils, and also for > > gdb. And we have a patch series from Romain Naour sitting in patchwork > > for several weeks. Right so I would think as of now the same patch should be applied to other affected instances of binutils and gdb. > > I don't know if Zlad's version is better or not than Romain's version. > > But at least Romain's version was handling all binutils and gdb > > versions, without patching directly Makefile.in files. So at first > > sight, it looked a lot better than Zlad's version. > > Ok, we can always revert if it isn't needed any more once Romain's > series is applied. In opposite Romain's fix only makes sense for BR and I don't really like that approach. Why implement hacks on top of upstream sources that are not upstreamable at all. Why not try "to fix" generic "missing" script if we do think it behaves improperly? -Alexey
Hi Peter, Thomas, On Tue, 2016-06-07 at 05:42 +0000, Alexey Brodkin wrote: > Hi Peter, Thomas, > > On Mon, 2016-06-06 at 23:34 +0200, Peter Korsgaard wrote: > > > > > > > > > > > > > > > > "Thomas" == Thomas Petazzoni <thomas.petazzoni@free-electrons.com> writes: > > > Hello, > > > On Mon, 06 Jun 2016 21:36:54 +0200, Peter Korsgaard wrote: > > > > >> > Signed-off-by: Zakharov Vlad <vzakhar@synopsys.com> > > >> > > >> Committed, thanks. > > > > > Why? > > > > It was fixing autobuilder issues and seemed like a sensible fix that > > could be upstreamed. Looking again, I do see that it patched Makefile.in > > and not Makefile.am, so that's not too nice though. > Indeed our goal was to fix an issue that causes all autobuilder jobs for ARC > to fail. The problem is host binutils couldn't be built any longer after we switched > to arc-2016.03 tools. And now we're unblocked. > > As for patching Makefile.in vs Makefile.am again that's just to do a minimal fix > in existing sources. Otherwise if we patch Makefile.am we'll need to regenerate > Makefile.in. > > Even though Vlad has already sent the same patch to Binutils mailing list > (https://sourceware.org/ml/binutils/2016-06/msg00053.html) that's indeed not > the right fix - he'll need to fix Makefile.am but at least we're moving > in upstreaming direction. > > > > > > This problem also exists for other versions of binutils, and also for > > > gdb. And we have a patch series from Romain Naour sitting in patchwork > > > for several weeks. > Right so I would think as of now the same patch should be applied to other > affected instances of binutils and gdb. > > > > > > I don't know if Zlad's version is better or not than Romain's version. > > > But at least Romain's version was handling all binutils and gdb > > > versions, without patching directly Makefile.in files. So at first > > > sight, it looked a lot better than Zlad's version. > > > > Ok, we can always revert if it isn't needed any more once Romain's > > series is applied. > > > In opposite Romain's fix only makes sense for BR and I don't really like that > approach. Why implement hacks on top of > upstream sources that are not > upstreamable at all. Why not try "to fix" generic "missing" script if we do think > it behaves improperly? So it looks like "makeinfo" is really required when building binutils/gdb from git sources, see that response in binutils mailing list: https://sourceware.org/ml/binutils/2016-06/msg00200.html So shall we just require "makeinfo" installed on build hosts? -Alexey
diff --git a/package/binutils/arc-2016.03/0800-Docs-Prevent-build-failures-when-makeinfo-is-missing.patch b/package/binutils/arc-2016.03/0800-Docs-Prevent-build-failures-when-makeinfo-is-missing.patch new file mode 100644 index 0000000..3f072de --- /dev/null +++ b/package/binutils/arc-2016.03/0800-Docs-Prevent-build-failures-when-makeinfo-is-missing.patch @@ -0,0 +1,159 @@ +From 5bd39e24e4903e49a0c301de4013c6b720862da1 Mon Sep 17 00:00:00 2001 +From: Zakharov Vlad <vzakhar@synopsys.com> +Date: Fri, 3 Jun 2016 21:33:43 +0300 +Subject: [PATCH] Docs: Prevent build failures when "makeinfo" is missing + +Build failed when "makeinfo" was missing on the build host. +This was happenning because "makeinfo" is required to build .info targets +and make exited with error. But missing "makeinfo" only stops us from +building docs ("missing" script already throws a warning on that regard). +Let's continue to build other targets. + +Now exit code of the script called "missing" is checked. +The value 127 means that "makeinfo" is not available on the build host. + +So when such value occurs 0 is returned to the top level makefile. +Documentation is not being built but further build of binutils continues. + +Signed-off-by: Zakharov Vlad <vzakhar@synopsys.com> +--- + bfd/doc/Makefile.in | 8 +++++++- + binutils/doc/Makefile.in | 8 +++++++- + gas/doc/Makefile.in | 8 +++++++- + gprof/Makefile.in | 8 +++++++- + ld/Makefile.in | 8 +++++++- + 5 files changed, 35 insertions(+), 5 deletions(-) + +diff --git a/bfd/doc/Makefile.in b/bfd/doc/Makefile.in +index be737f6..6c1eaf7 100644 +--- a/bfd/doc/Makefile.in ++++ b/bfd/doc/Makefile.in +@@ -468,6 +468,8 @@ mostlyclean-libtool: + clean-libtool: + -rm -rf .libs _libs + ++# Exit code 127 means that "makeinfo" is missing. So let's skip .info target ++# but return 0 to top level Makefile to build all other targets. + bfd.info: bfd.texinfo $(bfd_TEXINFOS) + restore=: && backupdir="$(am__leading_dot)am$$$$" && \ + rm -rf $$backupdir && mkdir $$backupdir && \ +@@ -481,7 +483,11 @@ bfd.info: bfd.texinfo $(bfd_TEXINFOS) + then \ + rc=0; \ + else \ +- rc=$$?; \ ++ if test $$? -eq 127; then \ ++ rc=0; \ ++ else \ ++ rc=$$?; \ ++ fi; \ + $$restore $$backupdir/* `echo "./$@" | sed 's|[^/]*$$||'`; \ + fi; \ + rm -rf $$backupdir; exit $$rc +diff --git a/binutils/doc/Makefile.in b/binutils/doc/Makefile.in +index ea3f938..8ed7884 100644 +--- a/binutils/doc/Makefile.in ++++ b/binutils/doc/Makefile.in +@@ -397,6 +397,8 @@ mostlyclean-libtool: + clean-libtool: + -rm -rf .libs _libs + ++# Exit code 127 means that "makeinfo" is missing. So let's skip .info target ++# but return 0 to top level Makefile to build all other targets. + binutils.info: binutils.texi + restore=: && backupdir="$(am__leading_dot)am$$$$" && \ + rm -rf $$backupdir && mkdir $$backupdir && \ +@@ -410,7 +412,11 @@ binutils.info: binutils.texi + then \ + rc=0; \ + else \ +- rc=$$?; \ ++ if test $$? -eq 127; then \ ++ rc=0; \ ++ else \ ++ rc=$$?; \ ++ fi; \ + $$restore $$backupdir/* `echo "./$@" | sed 's|[^/]*$$||'`; \ + fi; \ + rm -rf $$backupdir; exit $$rc +diff --git a/gas/doc/Makefile.in b/gas/doc/Makefile.in +index 9ca5e8e..3159bc3 100644 +--- a/gas/doc/Makefile.in ++++ b/gas/doc/Makefile.in +@@ -425,6 +425,8 @@ mostlyclean-libtool: + clean-libtool: + -rm -rf .libs _libs + ++# Exit code 127 means that "makeinfo" is missing. So let's skip .info target ++# but return 0 to top level Makefile to build all other targets. + as.info: as.texinfo $(as_TEXINFOS) + restore=: && backupdir="$(am__leading_dot)am$$$$" && \ + rm -rf $$backupdir && mkdir $$backupdir && \ +@@ -438,7 +440,11 @@ as.info: as.texinfo $(as_TEXINFOS) + then \ + rc=0; \ + else \ +- rc=$$?; \ ++ if test $$? -eq 127; then \ ++ rc=0; \ ++ else \ ++ rc=$$?; \ ++ fi; \ + $$restore $$backupdir/* `echo "./$@" | sed 's|[^/]*$$||'`; \ + fi; \ + rm -rf $$backupdir; exit $$rc +diff --git a/gprof/Makefile.in b/gprof/Makefile.in +index 67e85cd..c4040eb 100644 +--- a/gprof/Makefile.in ++++ b/gprof/Makefile.in +@@ -538,6 +538,8 @@ clean-libtool: + distclean-libtool: + -rm -f libtool config.lt + ++# Exit code 127 means that "makeinfo" is missing. So let's skip .info target ++# but return 0 to top level Makefile to build all other targets. + gprof.info: gprof.texi $(gprof_TEXINFOS) + restore=: && backupdir="$(am__leading_dot)am$$$$" && \ + rm -rf $$backupdir && mkdir $$backupdir && \ +@@ -551,7 +553,11 @@ gprof.info: gprof.texi $(gprof_TEXINFOS) + then \ + rc=0; \ + else \ +- rc=$$?; \ ++ if test $$? -eq 127; then \ ++ rc=0; \ ++ else \ ++ rc=$$?; \ ++ fi; \ + $$restore $$backupdir/* `echo "./$@" | sed 's|[^/]*$$||'`; \ + fi; \ + rm -rf $$backupdir; exit $$rc +diff --git a/ld/Makefile.in b/ld/Makefile.in +index 7c78198..08950a8 100644 +--- a/ld/Makefile.in ++++ b/ld/Makefile.in +@@ -1591,6 +1591,8 @@ clean-libtool: + distclean-libtool: + -rm -f libtool config.lt + ++# Exit code 127 means that "makeinfo" is missing. So let's skip .info target ++# but return 0 to top level Makefile to build all other targets. + ld.info: ld.texinfo $(ld_TEXINFOS) + restore=: && backupdir="$(am__leading_dot)am$$$$" && \ + rm -rf $$backupdir && mkdir $$backupdir && \ +@@ -1604,7 +1606,11 @@ ld.info: ld.texinfo $(ld_TEXINFOS) + then \ + rc=0; \ + else \ +- rc=$$?; \ ++ if test $$? -eq 127; then \ ++ rc=0; \ ++ else \ ++ rc=$$?; \ ++ fi; \ + $$restore $$backupdir/* `echo "./$@" | sed 's|[^/]*$$||'`; \ + fi; \ + rm -rf $$backupdir; exit $$rc +-- +2.5.5 +
Build failed when "makeinfo" was missing on the build host. This was happening because "makeinfo" is required to build .info targets and make exited with error. The issue hadn't appeared before as there was prebuilt documentation in ARC binutils tarballs, so no attempts had been made to build docs. Missing "makeinfo" only stops us from building docs ("missing" script already throws a warning on that regard). Let's continue to build other targets. Now exit code of the script called "missing" is checked. The value 127 means that "makeinfo" is not available on the build host. So when such value occurs, 0 is returned to the top level makefile. Documentation is not being built but further build of binutils continues. Fixes http://autobuild.buildroot.net/results/55c/55cd09a559016f4f252f0e4c27313b9806135cf4// Signed-off-by: Zakharov Vlad <vzakhar@synopsys.com> --- ...t-build-failures-when-makeinfo-is-missing.patch | 159 +++++++++++++++++++++ 1 file changed, 159 insertions(+) create mode 100644 package/binutils/arc-2016.03/0800-Docs-Prevent-build-failures-when-makeinfo-is-missing.patch