[Pkg-db-devel] Re: (MIPS|amd64)/gcc-assembly mutexes [#11575]

[Kurt Roeckx]

On Tue, Dec 07, 2004 at 05:02:56PM +0100, Thiemo Seufer wrote:
> 
> The appended patch (for db4.1) is derived from Kurt's one. The handling
> of --enable-pthreadsmutexes follows now the usual scheme (but is still
> disabled by default).

The db4.3 package (version 4.3.21-6) which had the proper changes
for x86_64/amd64 build succesfully last night with:
[...]
checking for mutexes... x86/gcc-assembly
[...]
cd obj && ../debian/runtests.tclsh || true
Modified test suite run started at: 23:57 12/06/04
Sleepycat Software: Berkeley DB 4.3.21: (December  5, 2004)
Running environment tests (23:57:38)
Running archive tests (23:57:48)
Running locking tests (23:58:03)
Running logging tests (00:00:57)
Running memory pool tests (00:03:00)
Running mutex tests (00:15:24)
Running transaction tests (00:17:48)
Running byte-order tests (00:22:32)
Running recno backing file tests (00:24:37)
Running DBM interface tests (00:24:38)
Running NDBM interface tests (00:24:38)
Running Hsearch interface tests (00:24:38)
grep ^FAIL obj/ALL.OUT || true
touch build


Note that it says it's using x86 (and not x86_64).  It seems the
original db4.3 sources already had:
#if (defined(i386) || defined(__i386__) || defined(__x86_64__)) && defined(__GNUC__)
For the i386 code, which also works on x86_64, it really does the
same anyway.  You might want to remove the
"|| defined(__x86_64__)" part.

I patched db4.1 (version 4.1.25-17) using Thiemo Seufer's patch,
removing the --enable-pthreadsmutexes call from the rules file,
and it build with x86_64/gcc-assembly.

I ran the full test and it report alot of errors in
some of tests, I think starting from the byte-order test.
I can provide the ALL.OUT if you want.


Kurt