Re: [Libunwind-devel] Plan for future releases?

[David Abdurachmanov]

> On 14 Feb 2016, at 10:03, David Abdurachmanov <address@hidden> wrote:
> 
> http://pkgs.fedoraproject.org/cgit/rpms/libunwind.git/tree/libunwind.spec
> http://pkgs.fedoraproject.org/cgit/rpms/libunwind.git/tree/

I decided to quickly scan the patches and check libunwind master.

If I didn't make a mistake:

## libunwind-1.1-fix-CVE-2015-3239.patch

Upstream, 396b6c7ab737e2bff244d640601c436a26260ca1

## libunwind-aarch64.patch

Full AArch64 support, this is upstreamed AFAIK.

## libunwind-arm-default-to-exidx.patch

RPM Changelog states:

* Tue Jun  2 2015 Tom Callaway <address@hidden> - 1.1-8 
- default arm unwinding method to exidx, old default of dwarf never works on 
Fedora
  (#1226806)

Here is the BZ item: https://bugzilla.redhat.com/show_bug.cgi?id=1226806#c4

I guess, same issue should affect other distributions, because usually debug 
information is separated from the binaries shipped in a separate package.

I checked debian (libunwind_1.1-3.2.debian.tar.xz) they don't do this.

Does not look like it's a bug in libunwind, and depend on TCMalloc.

But maybe it's reasonable to change the default? That's up to experts.
 
## libunwind-disable-setjmp.patch

The patch states:

At least x86_64 version cannot work, src/setjmp/setjmp.c and
src/setjmp/sigsetjmp.c are not even compiled, src/x86_64/longjmp.S does not
match src/setjmp/setjmp.c + include/tdep-x86_64/jmpbuf.h .

--enable-setjmp was introduced by 563b0ff3144c0e60a9aa63fdafb4dea66ecd3d50

Thus one should be able to disable setjmp without extra patches.

## libunwind-fix-ppc64_test_altivec.patch

Upstream, 4c62c4a9556f92495d93eddc7641497e1ce1e35c

- - -

Debian patches:

## 01-rpath.patch

Looks like distro specific.

They are disabling rpath on *-pc-linux-gnu hosts.

http://wiki.debian.org/RpathIssue

## 20123007_workaround_test-ptrace.patch

Looks to be not upstreamed, prevents some crash (segfault).

Patch is small, here it is:

Index: libunwind/tests/test-ptrace.c
===================================================================
--- libunwind.orig/tests/test-ptrace.c  2012-07-30 10:05:35.271168000 +0100
+++ libunwind/tests/test-ptrace.c       2012-07-30 11:39:37.309158623 +0100
@@ -182,6 +182,11 @@

       /* automated test case */
       argv = args;
+
+      /* Unless the args array is 'walked' the child
+         process is unable to access it and dies with a segfault */
+      fprintf(stderr, "Automated test (%s,%s,%s,%s)\n",
+              args[0],args[1],args[2],args[3]);
     }
   else if (argc > 1)
     while (argv[optind][0] == '-')


## 20130803-known_test_failure_to_XFAIL_TESTS.patch

Upstreamed, ae3dd9417a3d178cf51ce744d85b695dbf025b44

## 20130803-crasher_test_only_for_coredump.patch

Upstream, seems to be 61fad17d29ed661d15a80ed931c7822e5e7ea411

## 20140507-ppc_lib64.patch

Looks like this reverts a hunk from c340d04279f046b543066dbc8f36452602d7fd52

It avoids setting libdir='${exec_prefix}/lib64' on PPC64 target. This looks 
distro specific.

## ppc64el.diff

This one looks like powerpc64le-linux support, which was added by 
da0b1a146f5e6d3273a2719e9b11f3f91d077eb1

## AArch64-port.diff

Adding AArch64 support, upstreamed.

## fix-lzma-linking.patch

The patch look different, but it's upstreamed as 
dee708f4bc9aa169bd96e97c5e5ac851a10c7868

Cheers,
david