Bug 59394

Summary:	Building RPM 4.0.4 from Source (Available Mutex?)
Product:	[Retired] Red Hat Linux	Reporter:	Noam Rathaus <dolittle>
Component:	rpm	Assignee:	Jeff Johnson <jbj>
Status:	CLOSED WONTFIX	QA Contact:
Severity:	medium	Docs Contact:
Priority:	medium
Version:	7.2
Target Milestone:	---
Target Release:	---
Hardware:	strongarm
OS:	Linux
Whiteboard:
Fixed In Version:		Doc Type:	Bug Fix
Doc Text:		Story Points:	---
Clone Of:		Environment:
Last Closed:	2002-02-06 22:04:11 UTC	Type:	---
Regression:	---	Mount Type:	---
Documentation:	---	CRM:
Verified Versions:		Category:	---
oVirt Team:	---	RHEL 7.3 requirements from Atomic Host:
Cloudforms Team:	---	Target Upstream Version:
Embargoed:

Description Noam Rathaus 2002-02-06 22:04:06 UTC

From Bugzilla Helper:
User-Agent: Mozilla/4.0 (compatible; MSIE 5.5; Windows NT 5.0; T312461)

Description of problem:
Trying to compile the source pacakge results in a failure due to incorrect 
Mutex type. I am not able to find any working Mutex to use for the StrongARM 
architecutre.

Version-Release number of selected component (if applicable):


How reproducible:
Always

Steps to Reproduce:
1. Get the latest pacakge
2. Open it up in a directory
3. Using Intrinsync's StrongArm enviroment:

Run:
export CXX=gcc
export CC=gcc
export ac_cv_sizeof_char=1 \
ac_cv_sizeof_unsigned_char=1 \
ac_cv_sizeof_short=2 \
ac_cv_sizeof_unsigned_short=2 \
ac_cv_sizeof_int=4
export ac_cv_sizeof_long=4 \
ac_cv_sizeof_long_long=8 \
ac_cv_sizeof_unsigned_long_long=8 \
ac_cv_sizeof_float=4
export ac_cv_sizeof_double=8 \
db_cv_alignp_t=set \
db_cv_alignp_t="unsigned int"
export db_cv_fcntl_f_setfd=yes \
db_cv_sprintf_count=no \ 
db_cv_mutex="ia64/gcc-assembly"

(I tried, x86/gcc-assembly, ALPHA/gcc-assembly, ia64/gcc-assembly, and others)

export PATH=/root/Intrinsyc/arm-linux/bin:/root/Intrinsync/bin:$PATH
./configure --build=i386-linux --host=arm-linux --disable-aio 


Actual Results:  Different combinations of the same error:
/tmp/ccsmWE1E.s: Assembler messages:
/tmp/ccsmWE1E.s:85: Error: bad instruction `xchg1 r3=[r4,#0],r1'
make: *** [mut_tas.lo] Error 1

../mutex/mut_tas.c: In function `__db_tas_mutex_lock':
../mutex/mut_tas.c:109: inconsistent operand constraints in an `asm'

Expected Results:  Getting the binaries?! :)

Additional info:

I downloaded the latest stable and development package off www.rpm.org (4.0.4)

Comment 1 Jeff Johnson 2002-02-06 22:18:32 UTC

You're on an arm platform, you're gonna need arm
assembly voo-doo for a test-and-set mutex, no other platform's
assembly language will do.

Try looking in libpthread for arm specific mutexes, or ask an ARM
developer on an arm-specific mailing list for the Right Thing To Do.

Sorry, ARM is not a platform that's available to me, so I can't
undertake a fix flying blind, as it were. Please reopen this
bug if you have a patch for rpm on arm, however, I'll be
happy to integrate a known good patch

Comment 2 Noam Rathaus 2002-02-06 22:23:50 UTC

Wow, well Debian has been nicer with their reply, and they have one of their 
own, so I guess I will just move there.

If you will be kind enough to look into their's available here:
http://ftp.de.debian.org/debian/dists/woody/main/binary-arm/pool/main/r/rpm/

Comment 3 Noam Rathaus 2002-02-06 22:24:52 UTC

Wrong URL:
http://ftp.debian.org/debian/pool/main/r/rpm/

Comment 4 Jeff Johnson 2002-02-06 22:35:41 UTC

Thanks for the pointer. Here's what I find

+
+The rpm source also includes (but is not built using) a copy of db3, which has
+a BSD license (db/LICENSE in the source tree).
+

as it's db[34], not rpm, what needs the arm patch.

Comment 5 Jeff Johnson 2002-02-06 22:44:53 UTC

And I do not see an arm mutex patch at
../../d/db3/db3_3.2.9.patch.gz
(relative to the URL you've given me).

Again, if there's a patch, I'll be happy to integrate
into the db-4.0.14 code within rpm. I have no way of
testing still ...

Comment 6 Jeff Johnson 2002-02-06 23:57:03 UTC

Here's the arm asm voo-doo AFAICT:

However your path of least resistance is gonna be to
configure the internal db-4.0.14 with --enable-posixmutexes
to use the mutexes from libpthread.

Have fun with Debian however, I still cannot test the
above.


int
__pthread_spin_lock (pthread_spinlock_t *lock)
{
  unsigned int val;

  do
    asm volatile ("swp %0, %1, [%2]"
                  : "=r" (val)
                  : "0" (1), "r" (lock)
                  : "memory");
  while (val != 0);

  return 0;
}
weak_alias (__pthread_spin_lock, pthread_spin_lock)


int
__pthread_spin_trylock (pthread_spinlock_t *lock)
{
  unsigned int val;

  asm volatile ("swp %0, %1, [%2]"
                : "=r" (val)
                : "0" (1), "r" (lock)
                : "memory");

  return val ? EBUSY : 0;
}
weak_alias (__pthread_spin_trylock, pthread_spin_trylock)


int
__pthread_spin_unlock (pthread_spinlock_t *lock)
{
  return *lock = 0;
}
weak_alias (__pthread_spin_unlock, pthread_spin_unlock)