charm.git
19 years agoupdated.
Gengbin Zheng [Fri, 25 Oct 2002 16:59:20 +0000 (16:59 +0000)]
updated.

19 years agoIgnore error compiling f90 code.
Orion Lawlor [Fri, 25 Oct 2002 14:10:37 +0000 (14:10 +0000)]
Ignore error compiling f90 code.

19 years agoAggregate small outgoing messages using skt_sendV.
Orion Lawlor [Fri, 25 Oct 2002 00:07:56 +0000 (00:07 +0000)]
Aggregate small outgoing messages using skt_sendV.

19 years agoUse skt_sendV to aggregate small messages.
Orion Lawlor [Fri, 25 Oct 2002 00:06:33 +0000 (00:06 +0000)]
Use skt_sendV to aggregate small messages.

19 years agoUse skt_sendV to aggregate small messages, which dramatically
Orion Lawlor [Fri, 25 Oct 2002 00:05:53 +0000 (00:05 +0000)]
Use skt_sendV to aggregate small messages, which dramatically
improves the CCS ping time.

19 years agoAdded skt_sendV. Vector-style send (implemented here for small
Orion Lawlor [Fri, 25 Oct 2002 00:03:07 +0000 (00:03 +0000)]
Added skt_sendV.  Vector-style send (implemented here for small
messages using message combining) is a big, big win over TCP,
which delays the last of several small packets for up to 40ms,
in the hopes that more will come and be aggregated.

19 years agoChange parameter names to make it clear when we're casting
Orion Lawlor [Fri, 25 Oct 2002 00:01:28 +0000 (00:01 +0000)]
Change parameter names to make it clear when we're casting
a *converse* message directly to an envelope *, versus
when we should cast a *user* message using UsrToEnv.

19 years agoAdded "-fortran" link-time option, which allows the use
Orion Lawlor [Thu, 24 Oct 2002 22:37:51 +0000 (22:37 +0000)]
Added "-fortran" link-time option, which allows the use
of f90 libraries in Charm++ code.  In particular, this is
needed by the quantum chem. folks.

19 years agoAlways build conv-utilf.a (which is written in C++),
Orion Lawlor [Thu, 24 Oct 2002 22:37:08 +0000 (22:37 +0000)]
Always build conv-utilf.a (which is written in C++),
not just when you *know* you're linking fortran.

19 years agoRemoved make variable "$<", which on the DEC Alpha
Orion Lawlor [Thu, 24 Oct 2002 22:36:13 +0000 (22:36 +0000)]
Removed make variable "$<", which on the DEC Alpha
only works from a ".c.o"-style rule, not a regular rule.

19 years agoFixed makefile for elan-axp, which does not accept "$<"
Orion Lawlor [Thu, 24 Oct 2002 17:09:03 +0000 (17:09 +0000)]
Fixed makefile for elan-axp, which does not accept "$<"
as a valid variable.

19 years agoAdded tcharm and projections builds to top-level Makefile.
Terry L. Wilmarth [Wed, 23 Oct 2002 23:04:19 +0000 (23:04 +0000)]
Added tcharm and projections builds to top-level Makefile.

19 years agoExit with proper error code.
Orion Lawlor [Wed, 23 Oct 2002 21:11:29 +0000 (21:11 +0000)]
Exit with proper error code.

19 years agoFixed all Makefiles to point to the correct web dir path.
Terry L. Wilmarth [Wed, 23 Oct 2002 20:39:27 +0000 (20:39 +0000)]
Fixed all Makefiles to point to the correct web dir path.

19 years agoAdded -64 to f90 and f77 compilers, which just reinforces
Orion Lawlor [Wed, 23 Oct 2002 18:42:57 +0000 (18:42 +0000)]
Added -64 to f90 and f77 compilers, which just reinforces
the default on most origins, but actually changes it on some.

19 years agoAdded fixed-format f90 version; removed -W132 from normal f90 calls.
Orion Lawlor [Tue, 22 Oct 2002 22:22:30 +0000 (22:22 +0000)]
Added fixed-format f90 version; removed -W132 from normal f90 calls.
This prevents silly f90 warning "Fixed-format source option ignored;
free source mode selected".

19 years agoBugfix for gcc3--it instantiates the copy constructor.
Orion Lawlor [Tue, 22 Oct 2002 20:43:52 +0000 (20:43 +0000)]
Bugfix for gcc3--it instantiates the copy constructor.

19 years agoPrepare a short-time-limit batch job if run in interactive mode.
Orion Lawlor [Tue, 22 Oct 2002 17:34:40 +0000 (17:34 +0000)]
Prepare a short-time-limit batch job if run in interactive mode.
This is convenient for ordinary users, but absolutely necessary
for automated testing.

19 years agoAdded -LANG:std to allow use of <iostream> header.
Orion Lawlor [Tue, 22 Oct 2002 16:52:40 +0000 (16:52 +0000)]
Added -LANG:std to allow use of <iostream> header.

19 years agoCorrected spelling mistake in the name of Mahesh Karthikeyan
Rahul Joshi [Tue, 22 Oct 2002 02:52:28 +0000 (02:52 +0000)]
Corrected spelling mistake in the name of Mahesh Karthikeyan

19 years agoimmediate message for node group.
Gengbin Zheng [Tue, 22 Oct 2002 02:12:53 +0000 (02:12 +0000)]
immediate message for node group.

19 years agochanged prototype CmiPollImmediateMsg to CmiProbeImmediateMsg.
Gengbin Zheng [Tue, 22 Oct 2002 01:59:51 +0000 (01:59 +0000)]
changed prototype CmiPollImmediateMsg to CmiProbeImmediateMsg.

19 years agochanged CmiPollImmediateMsg to CmiProbeImmediateMsg cause it is not blocking.
Gengbin Zheng [Tue, 22 Oct 2002 01:59:03 +0000 (01:59 +0000)]
changed CmiPollImmediateMsg to CmiProbeImmediateMsg cause it is not blocking.

19 years agominor changes.
Gengbin Zheng [Mon, 21 Oct 2002 18:16:13 +0000 (18:16 +0000)]
minor changes.

19 years agoa short description of immediate message.
Gengbin Zheng [Mon, 21 Oct 2002 03:07:34 +0000 (03:07 +0000)]
a short description of immediate message.

19 years agochanged immediate message queue to PCQueue. and make sure the function to handle...
Gengbin Zheng [Mon, 21 Oct 2002 02:40:54 +0000 (02:40 +0000)]
changed immediate message queue to PCQueue. and make sure the function to handle the queue not to be reentrant..

19 years agoinit immediate bit to 0
Gengbin Zheng [Sun, 20 Oct 2002 22:52:40 +0000 (22:52 +0000)]
init immediate bit to 0

19 years agoset isImmediate for immediate messages entry.
Gengbin Zheng [Sun, 20 Oct 2002 21:50:00 +0000 (21:50 +0000)]
set isImmediate for immediate messages entry.

19 years agoadded is_immediate 1 bit field in envelop s_attrib. removed all special version of...
Gengbin Zheng [Sun, 20 Oct 2002 21:48:53 +0000 (21:48 +0000)]
added is_immediate 1 bit field in envelop s_attrib. removed all special version of ckSendImmediate.

19 years agoremoved the special version of CmiCommHandleMessage() for CmiHandleMessage() for...
Gengbin Zheng [Sun, 20 Oct 2002 21:42:31 +0000 (21:42 +0000)]
removed the special version of CmiCommHandleMessage() for CmiHandleMessage() for comm. thread.

19 years agobuild nodes_by_pe table for comm. threads. switch rank to dest rank in comm. thread...
Gengbin Zheng [Sun, 20 Oct 2002 21:38:31 +0000 (21:38 +0000)]
build nodes_by_pe table for comm. threads. switch rank to dest rank in comm. thread when executing an immediate handler.(as long as the immediate handler can be executed concurrently, this is safe)

19 years agocommunication thread has its pe set to Cmi_mynode+CmiNumPes() instead of -1.
Gengbin Zheng [Sun, 20 Oct 2002 21:34:03 +0000 (21:34 +0000)]
communication thread has its pe set to Cmi_mynode+CmiNumPes() instead of -1.

19 years ago*** empty log message ***
Sameer Kumar [Thu, 17 Oct 2002 23:08:03 +0000 (23:08 +0000)]
*** empty log message ***

19 years agoadding the communications library section into the manual
Sameer Kumar [Thu, 17 Oct 2002 23:07:43 +0000 (23:07 +0000)]
adding the communications library section into the manual

19 years agoUpdated for new Update_mesh call.
Orion Lawlor [Thu, 17 Oct 2002 23:02:23 +0000 (23:02 +0000)]
Updated for new Update_mesh call.

19 years agoFix for IBM SP-3: renamed "offsetof" (conflicts with C++ keyword
Orion Lawlor [Thu, 17 Oct 2002 23:01:07 +0000 (23:01 +0000)]
Fix for IBM SP-3: renamed "offsetof" (conflicts with C++ keyword
under lowercase) to "foffsetof".

19 years agoFixes for IBM SP-3. This machine's (stupid) f90 compiler just
Orion Lawlor [Thu, 17 Oct 2002 23:00:08 +0000 (23:00 +0000)]
Fixes for IBM SP-3.  This machine's (stupid) f90 compiler just
lowercases external names, which causes the f90 "init" and
C "init" to have the same name.

This means:
  -mesh_updated has to be passed in by the user, since we can't
statically know whether to call the C or F90 version; and they
both have the same names anyway.  This means a backward-compatability
breaking change to the arguments of FEM_Update_mesh.
  -"offsetof" can't be written in C (since it's a keyword); so
it has to be renamed "foffsetof".  This also breaks backward
compatability.

19 years agouse CmiCommHandleMessage instead of CmiHandleMessage.
Gengbin Zheng [Thu, 17 Oct 2002 17:58:58 +0000 (17:58 +0000)]
use CmiCommHandleMessage instead of CmiHandleMessage.

19 years agoclone CmiGetHandlerFunction to CmiCommGetHandlerFunction using CpvAccessOther, so...
Gengbin Zheng [Thu, 17 Oct 2002 17:58:27 +0000 (17:58 +0000)]
clone CmiGetHandlerFunction to CmiCommGetHandlerFunction using CpvAccessOther, so comm. thread can read the function table.

19 years agoAdded PUPable_abstract macro, which includes operator|
Orion Lawlor [Thu, 17 Oct 2002 16:47:15 +0000 (16:47 +0000)]
Added PUPable_abstract macro, which includes operator|
but no constructor registration.

19 years agofixed comments.
Gengbin Zheng [Thu, 17 Oct 2002 05:44:04 +0000 (05:44 +0000)]
fixed comments.

19 years agofixed comments // => /**/
Gengbin Zheng [Thu, 17 Oct 2002 05:42:48 +0000 (05:42 +0000)]
fixed comments // => /**/

19 years agoimmediate message support in Chare/Group/Array send.
Gengbin Zheng [Thu, 17 Oct 2002 05:23:23 +0000 (05:23 +0000)]
immediate message support in Chare/Group/Array send.

19 years agoimmediate message support.
Gengbin Zheng [Thu, 17 Oct 2002 05:17:11 +0000 (05:17 +0000)]
immediate message support.

19 years agofor [immediate] generate immediate send.
Gengbin Zheng [Thu, 17 Oct 2002 05:15:42 +0000 (05:15 +0000)]
for [immediate] generate immediate send.

19 years agosupport for immediate messages:
Gengbin Zheng [Thu, 17 Oct 2002 05:14:46 +0000 (05:14 +0000)]
support for immediate messages:
1. add data structure NodeState, which is supposed to include all Node level data structure(similar to CmiState), immediate message queue is in it.
2. CmiPollImmediateMsg() is exposed
3. set CMK_IMMEDIATE_MSG to 1 for net, mpi and elan version to enable it.

19 years agoUse .f90 for test f90 program.
Orion Lawlor [Wed, 16 Oct 2002 22:52:33 +0000 (22:52 +0000)]
Use .f90 for test f90 program.
Needed for MPI-SP.

19 years agoAdded two global variables:
Rahul Joshi [Wed, 16 Oct 2002 04:08:39 +0000 (04:08 +0000)]
Added two global variables:
1. dumpStep : the load balancing step at which to dump data
2. dumpFile: Name of the file into which to dump the data

Also added code to read these values from command line
parameters as: +LBDump <Step> +LBDumpFile <file name>

19 years agoAdded two global variables:
Rahul Joshi [Wed, 16 Oct 2002 04:07:01 +0000 (04:07 +0000)]
Added two global variables:
1. dumpStep : the load balancing step at which to dump data
2. dumpFile: Name of the file into which to dump the data

19 years agoAdded provision to dump the load balancing data, also fixed a bug in writeStatsMsg
Rahul Joshi [Wed, 16 Oct 2002 04:01:19 +0000 (04:01 +0000)]
Added provision to dump the load balancing data, also fixed a bug in writeStatsMsg

19 years agoplace holder for immediate messages: [immediate] attribute for entry.
Gengbin Zheng [Tue, 15 Oct 2002 03:49:18 +0000 (03:49 +0000)]
place holder for immediate messages: [immediate] attribute for entry.

19 years agoAdded prototype for exit().
Orion Lawlor [Mon, 14 Oct 2002 18:36:39 +0000 (18:36 +0000)]
Added prototype for exit().

19 years agoMinor source reorganization--moved internals into liveViz_impl.h,
Orion Lawlor [Sat, 12 Oct 2002 14:53:51 +0000 (14:53 +0000)]
Minor source reorganization--moved internals into liveViz_impl.h,
and created a separate liveVizPoll.ci file.

19 years agoThis checkin has the modifications to turn orion's liveViz into a split
Jonathan Booth [Fri, 11 Oct 2002 15:27:17 +0000 (15:27 +0000)]
This checkin has the modifications to turn orion's liveViz into a split
personality liveViz with both his original methods of running liveViz,
and the poll mode.

The code is messy and needs cleanup, but it is there and funcitonal. The
sample server (pollserver) has been developed and tested and checked in
as well.

This update does not contain anything more than the most basic polling mode,
none of the more advanced (so-called liveViz2, liveViz3) layers have been
implemented.

19 years agoadding bug fix!
Sameer Kumar [Thu, 10 Oct 2002 05:37:36 +0000 (05:37 +0000)]
adding bug fix!

19 years agonew version of comlibmanager which has learning
Sameer Kumar [Thu, 10 Oct 2002 05:10:07 +0000 (05:10 +0000)]
new version of comlibmanager which has learning

19 years agoput CthThreadTrace into the #else branch of all thread options, make it only seeable...
Gengbin Zheng [Wed, 9 Oct 2002 22:12:10 +0000 (22:12 +0000)]
put CthThreadTrace into the #else branch of all thread options, make it only seeable in qt threads. Moved it out.

19 years agoFixed backward compatibility bug
Chee Wai Lee [Tue, 8 Oct 2002 23:19:32 +0000 (23:19 +0000)]
Fixed backward compatibility bug

19 years agocorrect the option +tcharm_stacksize
Chao Huang [Tue, 8 Oct 2002 17:53:35 +0000 (17:53 +0000)]
correct the option +tcharm_stacksize

19 years agoAllowed mainchares to have 4 different constructors (), (void), (CkArgMsg *m)
Chee Wai Lee [Tue, 8 Oct 2002 04:15:36 +0000 (04:15 +0000)]
Allowed mainchares to have 4 different constructors (), (void), (CkArgMsg *m)
or (int argc, char **argv).

19 years agoAccidental check-in. Will not affect the rest of the charm system. Will fix
Chee Wai Lee [Tue, 8 Oct 2002 04:14:16 +0000 (04:14 +0000)]
Accidental check-in. Will not affect the rest of the charm system. Will fix
asap.

19 years agoMade the mainchares explicitly use CkArgMsg *.
Chee Wai Lee [Tue, 8 Oct 2002 04:12:54 +0000 (04:12 +0000)]
Made the mainchares explicitly use CkArgMsg *.

19 years agoAdded SIGTERM to list of signals we can do a sensible printout for.
Orion Lawlor [Mon, 7 Oct 2002 16:47:36 +0000 (16:47 +0000)]
Added SIGTERM to list of signals we can do a sensible printout for.

19 years agoUpdated capitalization of API calls.
Orion Lawlor [Mon, 7 Oct 2002 16:46:49 +0000 (16:46 +0000)]
Updated capitalization of API calls.

19 years agoserialno now starts from 1 for thread id.
Gengbin Zheng [Mon, 7 Oct 2002 03:00:30 +0000 (03:00 +0000)]
serialno now starts from 1 for thread id.

19 years agoFixed stupid bug-- when pup'ing, then calling destructor,
Gengbin Zheng [Sun, 6 Oct 2002 22:14:26 +0000 (22:14 +0000)]
Fixed stupid bug-- when pup'ing, then calling destructor,
the DDT arrays would be deleted twice.

19 years agoFixed bug-- when migrating, need to re-register worldPtr for
Gengbin Zheng [Sun, 6 Oct 2002 22:13:49 +0000 (22:13 +0000)]
Fixed bug-- when migrating, need to re-register worldPtr for
MPI_COMM_WORLD.

19 years agoa small fix for adjusting logentry.
Gengbin Zheng [Sun, 6 Oct 2002 01:04:24 +0000 (01:04 +0000)]
a small fix for adjusting logentry.

19 years agoEscaped {} in ghost sample.
Orion Lawlor [Sat, 5 Oct 2002 04:09:40 +0000 (04:09 +0000)]
Escaped {} in ghost sample.

19 years agoyet another fxi for LBDB is off.
Gengbin Zheng [Fri, 4 Oct 2002 15:33:24 +0000 (15:33 +0000)]
yet another fxi for LBDB is off.

19 years agoadded include memory.h for memcpy.
Gengbin Zheng [Fri, 4 Oct 2002 03:29:27 +0000 (03:29 +0000)]
added include memory.h for memcpy.

19 years ago+balancer option was only read by pe 0 and was kept on other pes. Fixed this.
Gengbin Zheng [Fri, 4 Oct 2002 01:31:21 +0000 (01:31 +0000)]
+balancer option was only read by pe 0 and was kept on other pes. Fixed this.

19 years agoCarefully ignore SIGPIPE *only* if we're inside a socket routine.
Orion Lawlor [Thu, 3 Oct 2002 20:07:53 +0000 (20:07 +0000)]
Carefully ignore SIGPIPE *only* if we're inside a socket routine.

Hopefully, this will do the right thing based on the skt_abort
routine.  That is, when:
  -A CCS client disconnects unexpectedly--right answer is to
just ignore the error.
  -Charmrun or a node unexpectly dies--right answer is to
exit properly.

19 years agoFinal version which seems to work!
Sameer Kumar [Wed, 2 Oct 2002 23:50:25 +0000 (23:50 +0000)]
Final version which seems to work!

19 years agofixing another bug
Sameer Kumar [Wed, 2 Oct 2002 21:42:55 +0000 (21:42 +0000)]
fixing another bug

19 years agofixed a bug
Sameer Kumar [Wed, 2 Oct 2002 21:33:28 +0000 (21:33 +0000)]
fixed a bug

19 years agoBug fix for destination of created object files.
Orion Lawlor [Wed, 2 Oct 2002 21:20:19 +0000 (21:20 +0000)]
Bug fix for destination of created object files.

For example,
gcc -c srcdir/foo.c
produces ./foo.o (NOT srcdir/foo.o), so
charmc -c srcdir/foo.c
should do the same thing, which means stripping off the
directory path from source files when deciding on .o names.

19 years agoadding new mesh
Sameer Kumar [Wed, 2 Oct 2002 20:28:55 +0000 (20:28 +0000)]
adding new mesh

19 years agoadding version with most optimal mesh.
Sameer Kumar [Wed, 2 Oct 2002 17:36:23 +0000 (17:36 +0000)]
adding version with most optimal mesh.

19 years agofixed a compile rerror in uth-linux when LBDB is off.
Gengbin Zheng [Wed, 2 Oct 2002 15:01:30 +0000 (15:01 +0000)]
fixed a compile rerror in uth-linux when LBDB is off.

19 years agoadded subsection to describe some load balancing utility functions.
Gengbin Zheng [Wed, 2 Oct 2002 06:19:37 +0000 (06:19 +0000)]
added subsection to describe some load balancing utility functions.

19 years agoexpose getLBDB().
Gengbin Zheng [Wed, 2 Oct 2002 05:41:04 +0000 (05:41 +0000)]
expose getLBDB().

19 years agoexpose setLBPeriod() function to set batsync period.
Gengbin Zheng [Wed, 2 Oct 2002 05:23:24 +0000 (05:23 +0000)]
expose setLBPeriod() function to set batsync period.

19 years agooops, my previous change about ResumeFromSync breaks no usesAtSync case. Fixed.
Gengbin Zheng [Tue, 1 Oct 2002 16:56:04 +0000 (16:56 +0000)]
oops, my previous change about ResumeFromSync breaks no usesAtSync case. Fixed.

19 years agowrite threadID and recv msg time(currenly not implemented) into projections log.
Gengbin Zheng [Tue, 1 Oct 2002 05:21:16 +0000 (05:21 +0000)]
write threadID and recv msg time(currenly not implemented) into projections log.

19 years agoadded threadID(which currently is 3 integers) to Converse thread struct.
Gengbin Zheng [Tue, 1 Oct 2002 05:18:33 +0000 (05:18 +0000)]
added threadID(which currently is 3 integers) to Converse thread struct.

19 years agoAdded MPI_Ireduce and sorted APIs in ampi.h
Chao Huang [Tue, 1 Oct 2002 03:17:20 +0000 (03:17 +0000)]
Added MPI_Ireduce and sorted APIs in ampi.h

19 years ago2 backward-compatability-breaking fixes:
Orion Lawlor [Mon, 30 Sep 2002 21:06:55 +0000 (21:06 +0000)]
2 backward-compatability-breaking fixes:

-NetFEM_Begin now takes a client number, to distinguish
between different virtual processors on the same processor.
It's not completely obvious this is better than just doing
a TCHARM_My_element, but it's certainly more flexible.

-The _Field routines are now capitalized _field, to be consistent
with the other frameworks.

19 years agoAdded test for signal-safe system() routine.
Orion Lawlor [Mon, 30 Sep 2002 21:05:01 +0000 (21:05 +0000)]
Added test for signal-safe system() routine.

19 years agoUse a signal-safe version of the "system" command.
Orion Lawlor [Mon, 30 Sep 2002 21:04:38 +0000 (21:04 +0000)]
Use a signal-safe version of the "system" command.

Without the signal-safe version, getting a SIGIO
in the middle of running the compiler just stops
the executution with EINTR.  SIGIO's are almost
inevitable in the net- version, so this is a good thing.

19 years agoAdded skt_server_ip routine, which both accepts and returns
Orion Lawlor [Mon, 30 Sep 2002 21:01:11 +0000 (21:01 +0000)]
Added skt_server_ip routine, which both accepts and returns
the IP address to bind to.  This turns out not to be as
useful as it looks, because (unlike a port number!) if you
pass in skt_invalid_ip, you get skt_invalid_ip back again.

The input value is still useful though, because on a multihomed
host it might be useful to bind to just one interface; or even
on a regular single-interface box you could demand it bind to
a specific IP, and get an error from bind if it's not what you
expect.

19 years agoMore support for array index [Max].
Jonathan Booth [Mon, 30 Sep 2002 19:54:46 +0000 (19:54 +0000)]
More support for array index [Max].

19 years agoUpdate defining CkArrayMax and similar for the array type [Max].
Jonathan Booth [Mon, 30 Sep 2002 19:52:47 +0000 (19:52 +0000)]
Update defining CkArrayMax and similar for the array type [Max].

19 years agoCorrected serious bug when packing readonly named "p".
Rahul Joshi [Mon, 30 Sep 2002 18:43:54 +0000 (18:43 +0000)]
Corrected serious bug when packing readonly named "p".

The old version created a PUP::er named p, and then did:
p|p;
which both totally ignores the readonly and (somehow) manages
to derail the pup routine, likely because PUP::er's aren't
pupable.

19 years ago*** empty log message ***
Chao Huang [Mon, 30 Sep 2002 03:45:19 +0000 (03:45 +0000)]
*** empty log message ***

19 years agoAdding the period as a parameter to the strategy!
Sameer Kumar [Mon, 30 Sep 2002 01:46:19 +0000 (01:46 +0000)]
Adding the period as a parameter to the strategy!

19 years agoAdding a streaming strategy!
Sameer Kumar [Mon, 30 Sep 2002 01:11:35 +0000 (01:11 +0000)]
Adding a streaming strategy!

19 years ago*** empty log message ***
Chao Huang [Sun, 29 Sep 2002 22:21:18 +0000 (22:21 +0000)]
*** empty log message ***

19 years agoCommiting with a bug fix for all processors being used!
Sameer Kumar [Sun, 29 Sep 2002 19:37:34 +0000 (19:37 +0000)]
Commiting with a bug fix for all processors being used!

19 years agoconfigure broken in net-win32. somehow $? of tests is lost when enters test_result...
Gengbin Zheng [Sun, 29 Sep 2002 18:05:03 +0000 (18:05 +0000)]
configure broken in net-win32. somehow $? of tests is lost when enters test_result. I have to make $? an argument of test_result().