microsoft / omi Goto Github PK

There are a number of issues to be fixed:

1. Server isn't always given enough time to properly shut down, causing next tests to fail.
2. If auth tests are skipped, it should show as failed (except on Travis).
3. Enable tests to run using either "regress" or "make test".
4. If server is running, shut it down before doing unit-tests.
5. Require that an account with valid credential be created before running tests, and that environment variables OMI_USER and OMI_PASSWORD be set to the credential.
6. "sudo -E" is not supported on some platforms.

When running authentication unit-tests, the server must be run as super-user. This creates some unique problems in starting up and shutting down the omiserver.

For authentication tests, omiserver must be started via the sudo command. Since sudo doesn't seem to work well with system(3) command, one must start up omiserver using fork/exec. And since the start up and shutting down of omiserver both require fork/exec, you now have an asynchronous situation where the next test may start before the previous one shuts down cleanly. And since both tests rely on the same pid file, you may have situations where the pid file of second test gets deleted by first test.

We've gotten around the problem so far by putting in more and more sleep. This is both unsafe and unreliable. A better way is needed.

Good evening. I've been trying to create a PowerShell cmdlet for OMI from CDXML on Linux without success. I've been through Richard Siddaway's book and blog which describe how to do this on Windows, but when I go to run my shiny new cmdlet on Linux, I consistently get back the error below. Any chance I can sweet talk you into producing an example of a working cdxml cmdlet for PowerShell on Linux that I can use to Get-XYZ_Frog ?

I feel like there is a lot of potential here. I just need help getting started. Thanks for your consideration of the matter.

"Exception calling 'ProcessRecord' with '1' argument(s): 'The method or operation is not implemented.'"
At line: 108 char:5

• $__cmdletization_objectModeWrapper.ProcessRecord($__cmdletizatio ..."
• CategoryInfo: NotSpecified: (:) [], ParentContainsErrorRecordException
• FullyQualifiedError ID : NotImplementedException

The reason is the failure to end the strand. If an enumerate was started an enuerate context is still active and the EnumeratePullRequest tries to inspect the not-existing headers.

The solution is to deactivate the enumerate context and post a bad result to the strand.

on hp_v3_ia64, libtest_wsman.so:TestWSMAN_GetClass_CIM_Error_StandardRtti and libtest_wsman.so:TestWSMAN_GetClass_CIM_Error_StandardRtti_OOP tests fail consistently.

This is the actual response from server:

<SOAP-ENV:Envelope xmlns:SOAP-ENV="http://www.w3.org/2003/05/soap-envelope" xmlns:wsa="http://schemas.xmlsoap.org/ws/2004/08/addressing" xmlns:wsen="http://schemas.xmlsoap.org/ws/2004/09/enumeration" xmlns:e="http://schemas.xmlsoap.org/ws/2004/08/eventing" xmlns:xsi="http://www.w3.org/2001/XMLSchema-instance" xmlns:wsmb="http://schemas.dmtf.org/wbem/wsman/1/cimbinding.xsd" xmlns:wsman="http://schemas.dmtf.org/wbem/wsman/1/wsman.xsd" xmlns:wxf="http://schemas.xmlsoap.org/ws/2004/09/transfer" xmlns:cim="http://schemas.dmtf.org/wbem/wscim/1/common" xmlns:msftwinrm="http://schemas.microsoft.com/wbem/wsman/1/wsman.xsd" xmlns:wsmid="http://schemas.dmtf.org/wbem/wsman/identity/1/wsmanidentity.xsd">
SOAP-ENV:Header
wsa:To
http://schemas.xmlsoap.org/ws/2004/08/addressing/role/anonymous
</wsa:To>
wsa:Action
http://schemas.xmlsoap.org/ws/2004/09/transfer/GetResponse
</wsa:Action>
wsa:MessageIDuuid:2D9060DE-4533-0005-0000-000000010000</wsa:MessageID>
wsa:RelatesTouuid:DFCFFC30-89CA-4C3F-8E8C-4DA843C58615</wsa:RelatesTo>
</SOAP-ENV:Header>
SOAP-ENV:Body


2.22.1


CIM::Interop


true






true



<PROPERTY.ARRAY NAME="MessageArguments" TYPE="string" CLASSORIGIN="CIM_Error"/>



<PROPERTY.ARRAY NAME="RecommendedActions" TYPE="string" CLASSORIGIN="CIM_Error"/>


0





</SOAP-ENV:Body>
</SOAP-ENV:Envelope>

Supported platforms include all PowerShell-supported platforms, plus Linux platforms.

tests/codec/mof/blue/consts.c

test case "test13" must be deactivated on solaris, hpux, aix due to limitations of their strtoull
which cannot parse the number "0x7fffffffffffffff" . They all share the same bug that they do not overrange, but do not perform the conversion. Macos and 64 bit Linux are able to handle it only because they use a 128 bit integer type to accumulate in the function.

When errors are encountered at parse time deserializing class or instance MOFs, the errors reported by the parser are lost. This is due to MI_RESULT_TYPE_MOF_PARSER not being handled by OMI_ErrorFromErrorCode.

this is seen with DSC/Linux when errors are encountered parsing a configuration or meta MOF that contains errors. Instead of receiving a specific error, a generalized file load error is reported.

See Dsc/Linux issue microsoft/PowerShell-DSC-for-Linux#83 for an example.

The configure script copies some base header files and changes them based on the build configuration.

This is often problematic in that, after doing a rebase operation, builds may start to fail until configure is rerun. This is not something that developers expect, wasting a potentially large amount of time.

Can this be resolved such that builds are more stable without requiring configure to modify base files?

On Ubuntu 16.10, it became clear that readdir_r is being deprecated. See the manpage for readdir_r (or see www manpage for further details.

Basically, while readdir_r was supposed to be thread-safe (and is), it has a host of problems with the interface. Future plans are to make readdir thread-safe (possibly through external synchronization), and suggest that people use that instead.

Note that use of readdir_r, by default, triggers a compiler warning on Ubuntu 16.10.

Note that there are actually TWO readdir_r functions:

1. readdir_r (section 3 of manpage), that has the problems specified above, and
2. readdir_r (section 2 of manpage), that isn't intended to be called directly, and is only reachable via the _syscall3 interface - and there is no glibc convenience function for this). Note that readdir_r (2) has been superseded by getdents (2), which allows getting multiple directory entries in one call.

Because of our universal images, we'll likely need to define our own version of readdir that we guarantee to be thread safe that basically re-implements readdir_r (3) using underlying readdir_r (2) or getdents (2) interfaces. We will need to test such an interface via a universal binary and verify that it runs on all of the Linux platforms that we care about, even when built on an old system.

The Page 1 is missing from the PDF that explains what OMI is.

I'm not sure yet what is going on here, but I'll report it anyway with what I know.

I see a steady growth in resident memory used by the omiagent process on Ubuntu 16.04.1 LTS VM:s in Azure running Docker 1.12.3 together with the latest OmsAgent, v1.2.0-148.

Attached is an image showing the result in OMS after a systemctl restart omid, freeing up about 300Mb of memory per node, then starting growing linearly again.

The growth is identical on all docker hosts, regardless of their running containers which are different (apart from one global service). The graph shows five swarm members with the same level of growth from different base levels. Note that the graph, though hard to see, also includes one management node not running the docker daemon (the flat dark green line), and it does not show this growth. The management VM is smaller (A1 vs A2), but otherwise runs the same software apart from the Docker daemon.

I cannot find anything in any logs seeming related.

$ uname -a
Linux kth-integral-2 4.4.0-47-generic #68-Ubuntu SMP Wed Oct 26 19:39:52 UTC 2016 x86_64 x86_64 x86_64 GNU/Linux

$ sudo docker version
Client:
Version: 1.12.3
API version: 1.24
Go version: go1.6.3
Git commit: 6b644ec
Built: Wed Oct 26 22:01:48 2016
OS/Arch: linux/amd64

Server:
Version: 1.12.3
API version: 1.24
Go version: go1.6.3
Git commit: 6b644ec
Built: Wed Oct 26 22:01:48 2016
OS/Arch: linux/amd64

it should be set to the target hostname, In fact Windows puts the target host:port

currently there is one per request which makes things really heavyweight, and complicates shutdown of each request significantly making it very error prone

OMI history has been rewritten. This is super invasive.

The easiest way to resolve this is to simply reclone your repository. This is the safest way to make sure that you don't inadvertently push the "bad history" back to GitHub.

Let me know if you have any questions.

@Microsoft/omi-devs
@Microsoft/ostc-devs
@Microsoft/omsagent-devs

This issue will be deleted before we go public.

windows has the setting NetworkDelayms = 5000 which it adds to the operation timeout for the wait time on the http request before timing out

Hi folks,

I needed to change how OMI is accessed due to additional team members; I needed to change the granularity of access.

We have an additional team now:

1. @MSFTOSSMgmt/omsdevs : Controls access to all repositories except OMI
2. @MSFTOSSMgmt/omidevs : Controls access to the OMI repository

If you are doing active development in OMI, then you need to be in the OMIDevs team. If you only need read access to to OMI project, then just being in OMSDevs should be sufficient (which everyone getting this is except for JohnA and Andy).

Bottom line: This shouldn't affect anyone except the few of us directly contributing to the OMI project.

If you have any problems, let me know, and I'll take care of it.

Note: I would anticipate that, as time progresses, more granular teams will be set up. In all cases, the OMSDevs team will retain access to the repositories that they need, although some may become read-only.

/Jeff

I need to make packages for OMI per our internal project, and while the install builder setup is available within OMI, as far as I and @jeffaco can tell, the OMI packages can't be built from OMI itself.

@johnkord are we mistaken here, or do we need to get the installbuilder scripts made available in OMI (or a separate repo)?

We should get a README.md written for OMI, covering at least what the project is, and how to build (packages and developer), test, and run it.

@KrisBash @jeffaco @paulcallen

The first row of the table in the "Configuring OMI Server" section has "comma-seperated" instead of "comma-separated"

A number of regress test cases seem to be timing out after a 24 second wait.

A partial list would be :
libtest_protocol.so:TestTransferingEnumerateInstanceRequest 24.0015
libtest_protocol.so:TestTransferingResult 24.0016
libtest_protocol.so:TestTransferingGetInstanceReqt 24.0026
libtest_protocol.so:TestTransferingPostInstance 24.0021
libtest_protocol.so:TestTransferingInvoke 24.0017

PowerShell remote management of office 365 requires this feature to work. Typically you are given I high-level end-point to connect to and then that redirects to a user or organization specific end-point on a different machine or end-point.

we have an internal mi_result that should not be exposed directly to the client. instead we need to create a CIM_Error object with the correct category stating it is a timeout, and the actual mi_result will be an mi_result_failed.
NOTE: When this changes the powershell remoting client will need to be updated otherwise it will be completely broken with its Receive processing which relies on Timeout requests from the server.

• Client-side wsman timeout

• Invalid machine

• Bad credentials

• Bad options

OMI (like many other programs) requires the Homebrew version of OpenSSL. We used to follow .NET Core's instructions and brew link --force the libraries into the system locations. This was, of course, a terrible idea and Homebrew/brew#597 prevents it from being done (for security reasons), which caused our CI build failures for OS X.

./omicli gi --hostname localhost oop/requestor/test/cpp { MSFT_President Key 1 }

returns:

instance of MSFT_President
{
Key=1
First=George
Last=Washington
}

Debugger shows that it's using the credential of 'root'

Tracking here so the issue doesn't get lost.

@agup006 Last I heard this was being discussed with legal. Has there been any movement?

While OMI is "open-source" the only release is through The Open Group, and so our real repository is private.

This causes a lot of pain with our other internal project which takes a dependency on OMI, as every user needs to get authenticated in both GitHub organizations.

omicli always crashes on this request for me:

$ omicli -t -xml -u ... -p ... --hostname ... gi root/cimv2 '{' Win32_Service name spooler '}'

The crash happens here when accessing the "_encode" array outside of its bounds:

Program terminated with signal SIGSEGV, Segmentation fault.
#0 0x000000000040fd3b in WriteBuffer_Char (clientBuffer=0x1df59f0 "<INSTANCE CLASSNAME="Win32_Service"><PROPERTY NAME="AcceptPause" TYPE="string">false<PROPERTY NAME="AcceptStop" TYPE="string">true<PROPERTY NAME="Ca"..., clientBufferLength=10000, clientBufferNeeded=0x7ffccb193ecc, charToWrite=-30 '\342', escapingDepth=1, result=0x7ffccb193e18) at xmlserializer.c:894
894 (_encode[(unsigned int)charToWrite].str == NULL))

Here's the interesting part of the received string - the problematic character being E2:

$ tail -1 omiclient-recv.trc | xxd
...
0000490: 7365 7276 6963 652c 2079 6f75 2077 6f6e service, you won
00004a0: e280 9974 2062 6520 6162 6c65 2074 6f20 ...t be able to
00004b0: 7072 696e 7420 6f72 2073 6565 2079 6f75 print or see you
00004c0: 7220 7072 696e 7465 7273 2e3c 2f70 3a44 r printers.</p:D
...

The fix is simple - just ensure array index is OK before accessing it.

Installed the newest version, omsagent-1.2.0-148, of the oms agent seem to crash continuously and fills upp the disk with core dumps under the following path, /var/opt/omi/run/.
The issue seems to be with the omi provided in that package and verified that it crashes on ubuntu 16.04.1 and centos 7.2. Previous versions of the agent worked without issues.

Running on the vm's is
Docker 1.12.3
Omsagent 1.2.0-148 (omi 1.1.0)

We get the following errors on ubuntu/centos machines

[root@myhost run]# journalctl -f | grep omiag
Nov 23 14:40:22 myhost kernel: omiagent[57281]: segfault at 10 ip 00007f12d7b92869 sp 00007fffdab230e0 error 4 in libcontainer.so[7f12d7b33000+77000]
Nov 23 14:40:59 myhost kernel: omiagent[57322]: segfault at 10 ip 00007f03bc34f869 sp 00007ffd14f432a0 error 4 in libcontainer.so[7f03bc2f0000+77000]
Nov 23 14:41:36 myhost kernel: omiagent[57399]: segfault at 10 ip 00007fe3fedf4869 sp 00007ffc5f19ede0 error 4 in libcontainer.so[7fe3fed95000+77000]
Nov 23 14:42:13 myhost kernel: omiagent[57427]: segfault at 10 ip 00007f3d70af4869 sp 00007ffcf7b88350 error 4 in libcontainer.so[7f3d70a95000+77000]

A collegue ran a gdb trace and i've attached it to the issue
gdb.txt

User.c has the logic to check the username/password. The check for the return value is wrong, thus it causes accidentally returning the last username/password when the username/password does not exist in the system.

The resource URI can change, for instance managing Office 365 uses non-standard resource URI, so we need to use an alternative operation option to identify it,
Invalid shell packets are generated in the resource URI changes so this is a must fix.

This test fails on aix and hpux . It appears to be related to the parser callback needsQualifier. The problem appears to be isolated to the test case and not underlying product code.

Downloaded a clean Ubuntu 16 server.
Dependencies for compiling should be listed including libraries (they are not)
C++
cmake (to start - seems obvious but it isn't)

Fixed the initial reporting of an error. Now we need to report the correct error and signal that it was a client timeout, not server.

test_pal.c,pp test case test multibytecharacterlength3 fails on aix. The utf8length of the function is returned as zero.

This case has been deactivated for aix. It is deactivated for windows already.

It's been observed that sometimes omiserver from previous pbuild runs do not get shutdown properly, and thus interfere with new pbuild runs.

Make sure pid files are cleaned up before unittests.

only user is encoded in the auth header rather than full user@domain.
basic auth should use the username as is with no changes.

Need to do proper channel binding for kerberos authorisation. Currently we are assuming spnego resulting in ntlm, and channel binding is null, but in kerberos we need to bind to the channel

On CentOS7, basic-auth unit-tests fail. Investigations show that this is because the server tries to spawn omiagent and fails. Omiagent is invoked using the test account credentials (omi_test), which does not have access to the omiagent executable because my /home/v-geflem directory is by default 700 on CentOS7. In most other systems, the default permission is 755 for home directories.

Temporarily change home directory to 711, and the tests run cleanly.

error output:

test_timer.cpp(169):
Assert: timerTestStrand.timeoutTime-startTimeUsec >= FIFTY_MILLISECONDS_AS_USEC
(Timeout took shorter than expected)
[Failed] libtest_base.so:TimerTest_BasicTimeout_Success 0.054719

Investigations show that the actual timeout was sometimes 49.* msecs.

Although various platforms provide high-precision timers, there isn't an easy way to implement this on all platforms. See http://stackoverflow.com/questions/361363/how-to-measure-time-in-milliseconds-using-ansi-c.

Our implementation is located in pal/sleep.c. On most platforms, _POSIX_MONOTONIC_CLOCK does not appear to be > 0, so just basic gettimeofday() is used. It's not trivial to just enable clock_gettime, as many older platforms also require linking with -lrt library.

On solaris, a test case failed at Assertion failed: !info->otherAckPending, file strand.c, line 547.

The failure is where a request is submitted before the ack is received for another request.

gzipped tar file attached with the core and binaries.

code for reference:

(dbx) list 530,560
530 DEBUG_ASSERT( NULL != info->userFT->PostControl );
531 DEBUG_ASSERT( NULL != info->stored.controlMsg );
532
533 info->userFT->PostControl( self, info->stored.controlMsg );
534 Message_Release( info->stored.controlMsg ); // now we can remove the reference added on StrandInteraction*
535
536 return MI_FALSE;
537 }
538
539 //------------------------------------------------------------------------------------------------------------
540 MI_INLINE
541 MI_Boolean _StrandMethodImp_Post( In Strand* self, In InteractionInfo* info)
542 {
543 DEBUG_ASSERT( info->opened );
544 DEBUG_ASSERT( !info->otherClosedThis );
545 DEBUG_ASSERT( NULL != info->userFT->Post );
546 DEBUG_ASSERT( NULL != info->stored.msg );
547 DEBUG_ASSERT( !info->otherAckPending );
548
549 info->otherAckPending = MI_TRUE;
550 info->userFT->Post( self, info->stored.msg );
551 Message_Release( info->stored.msg ); // now we can remove the reference added on StrandInteraction*
552
553 return MI_FALSE;
554 }
555
556 //------------------------------------------------------------------------------------------------------------
557 MI_INLINE
558 MI_Boolean _StrandMethodImp_Ack( In Strand* self, In InteractionInfo* info)
559 {
560 // note that info->otherClosedThis may be true here

The thread stacks are 👍

(dbx) where
current thread: t@80
[1] _lwp_kill(0x0, 0x15f84aa0, 0x0, 0x6, 0x0, 0x0), at 0xf3349fa0
[2] raise(0x6, 0xffffffef, 0xffffffec, 0xf9bfaf90, 0x5, 0x6), at 0xf32f3338
[3] abort(0xf9bfb090, 0xf33ee1f0, 0x6, 0x1, 0xf33efa80, 0x0), at 0xf32cac6c
[4] _assert(0xfe386ee4, 0xfe3924a8, 0x223, 0x218aa8, 0xf9bfb090, 0xf33ee1f0), at 0xf32caf28
=>[5] _StrandMethodImp_Post(self = 0x10f744, info = 0x10f76c), line 547 in "strand.c"
[6] _StrandMethod_Post(self = 0x10f744), line 692 in "strand.c"
[7] _Strand_ExecuteLoop(self = 0x10f744, state = 577), line 2874 in "strand.c"
[8] _Strand_ScheduleImp(self = 0x10f744, methodBit = 64U, allowMultiSchedule = '\0', fromStrand = (nil), entryOperationBit = 0), line 3036 in "strand.c"
[9] _StrandInteraction_Post(interaction = 0x10f76c, msg = 0x218aa8), line 363 in "strand.c"
[10] _Strand_PostAndLeaveStrand_Imp(strand = 0x221a88, info = 0x221ab0, msg = 0x218aa8), line 555 in "Strand.h"
[11] Strand_PostAndLeaveStrand(strand = 0x221a88, msg = 0x218aa8), line 1061 in "Strand.h"
[12] ProtocolSocket_Aux_PostMsg(self = 0x221a88), line 353 in "protocol.c"
[13] _StrandMethod_Aux0(self = 0x221a88), line 747 in "strand.c"
[14] _Strand_ExecuteLoop(self = 0x221a88, state = 4097), line 2874 in "strand.c"
[15] _Strand_ScheduleImp(self = 0x221a88, methodBit = 4096U, allowMultiSchedule = '\0', fromStrand = (nil), entryOperationBit = 0), line 3036 in "strand.c"
[16] Strand_ScheduleAux(self = 0x221a88, auxMethodNumber = 0), line 1733 in "Strand.h"
[17] _ProcessReceivedMessage(handler = 0x221a68), line 1015 in "protocol.c"
[18] _ReadAllPages(handler = 0x221a68), line 1194 in "protocol.c"
[19] _RequestCallbackRead(handler = 0x221a68), line 1216 in "protocol.c"
[20] _RequestCallback(sel = 0x221f50, handlerIn = 0x221a68, mask = 2U, currentTimeUsec = 7783670730051ULL), line 1240 in "protocol.c"
[21] Selector_Run(self = 0x221f50, timeoutUsec = 0, noReadsMo(dbx) where t@79
current thread: t@79
=>[1] __lwp_park(0x4, 0x0, 0x0, 0x1, 0x0, 0xfed90080), at 0xf3345478
[2] cond_sleep_queue(0x2f918, 0x2f928, 0x0, 0x0, 0x0, 0x0), at 0xf333f020
[3] cond_wait_queue(0x2f918, 0x2f928, 0x0, 0x0, 0x0, 0x0), at 0xf333f288
[4] cond_wait(0x2f918, 0x2f928, 0x0, 0x0, 0xfd5d1240, 0x0), at 0xf333f8d0
[5] pthread_cond_wait(0x2f918, 0x2f928, 0x0, 0x0, 0x2f928, 0xfffc00), at 0xf333f90c
[6] CondLock_Wait(key = 1161952, destination = 0x11bb18, comparand = 0, spinCount = 4294967295U), line 1199 in "condlockposix.c"
[7] InteractionProtocolHandler_ThreadShutdown(_application = 0x11bae0), line 2435 in "InteractionProtocolHandler.c"
[8] _Wrapper(param = 0x115a58), line 221 in "thread.c"

(dbx) thread -info t@1
Thread t@1 (0xfeda2a40) at priority 0
state: asleep on _slots+0x5588 (0x32a18): thread condition variable
base function: 0x0: 0x00000000() stack: 0xffc00000[8388608]
flags: (none)
masked signals: (none)
Currently inactive in __lwp_park
(dbx) where t@1
current thread: t@1
=>[1] __lwp_park(0x4, 0x0, 0x0, 0x1, 0x0, 0xfed90000), at 0xf3345478
[2] cond_sleep_queue(0x32a18, 0x32a28, 0x0, 0x0, 0x0, 0x10f744), at 0xf333f020
[3] cond_wait_queue(0x32a18, 0x32a28, 0x0, 0x0, 0x0, 0x0), at 0xf333f288
[4] cond_wait(0x32a18, 0x32a28, 0x0, 0x0, 0xfeda2a40, 0x156f70), at 0xf333f8d0
[5] pthread_cond_wait(0x32a18, 0x32a28, 0x0, 0x0, 0x32a28, 0xf33ee1f0), at 0xf333f90c
[6] CondLock_Wait(key = 1111616, destination = 0x10f6e4, comparand = 0, spinCount = 4294967295U), line 1199 in "condlockposix.c"
[7] Operation_GetInstance_Result(operation = 0xffbfdaa8, instance = 0xffbfd9dc, moreResults = 0xffbfd9e3 "", result = 0xffbfd9e8, errorMessage = 0xffbfd9d4, completionDetails = 0xffbfd9d0), line 2910 in "Operation.c"
[8] MI_Operation_GetInstance(operation = 0xffbfdaa8, instance = 0xffbfd9dc, moreResults = 0xffbfd9e3 "", result = 0xffbfd9e8, errorMessage = 0xffbfd9d4, completionDetails = 0xffbfd9d0), line 7723 in "MI.h"
[9] ConsumeInstanceResults(miOperation = 0xffbfdaa8), line 651 in "cli_c.c"
[10] Associators(miSession = 0xffbfdb78, argc = 5, argv = 0x1153bc), line 1514 in "cli_c.c"
[11] climain(argc = 8, argv = 0x1153b0), line 2768 in "cli_c.c"
[12] Exec(cmd = 0xf98fa884 "omicli -synchronous a test/cpp { X_numberWorld Name theWorld }", out = CLASS, err = CLASS), line 370 in "test_cli.cpp"
[13] TestOMICLI14_Sync(_NitsContext = 0x115638), line 768 in "test_cli.cpp"
[14] NitsBody_NewInterfaceTest(test = CLASS), line 1159 in "Switch.cpp"
[15] TestSystem::Test::RunBody(this = 0xf9a75e10), line 1045 in "Switch.cpp"
[16] TestSystem::Test::ContinueProcessingAfterSetup(this = 0xf9a75e10), line 958 in "Switch.cpp"
[17] NitsRunContinuation(test = CLASS), line 1412 in "Switch.cpp"
[18] TestCliSetup(_NitsContext = 0x115678), line 385 in "test_cli.cpp"
[19] NitsBody_NewInterfaceTest(test = CLASS), line 1159 in "Switch.cpp"
[20] NitsSetup_NewInterfaceTest(test = CLASS), line 1235 in "Switch.cpp"
[21] TestSystem::Switch::RunSetup(this = 0xf9a75e10), line 283 in "Switch.cpp"
[22] TestSystem::Test::RunVariation(this = 0xf9a75e10), line 1007 in "Switch.cpp"
[23] TestSystem::Test::Execute(this = 0xf9a75e10, choices = (nil)), line 558 in "Switch.cpp"
[24] TestSystem::Run::ExecuteTest(this = 0xff2da314, test = 0xf9a75e10, choices = (nil), executeIfTypeIs = BodyFixture), line 988 in "Run.cpp"
[25] TestSystem::Module::RunTests(this = 0x1170f8, start = 0), line 286 in "Run.cpp"
[26] TestSystem::Module::Execute(this = 0x1170f8), line 352 in "Run.cpp"
[27] TestSystem::Run::Execute(this = 0xff2da314), line 750 in "Run.cpp"
[28] NitsExecuteRun(), line 38 in "Run.cpp"
[29] main(argc = 2, argv = 0xffbff8a4), line 151 in "nits.cpp"
(dbx) where t@2
current thread: t@2
=>[1] __nanosleep(0x4, 0x0, 0x0, 0x1, 0x0, 0x0), at 0xf334936c
[2] nanosleep(0xfc35bd70, 0x0, 0x0, 0x0, 0xfd5d0240, 0x0), at 0xf3335874
[3] Sleep_Milliseconds(milliseconds = 1U), line 31 in "sleep.h"
[4] __TimedWaitHelper(sem = 0xfece0000, milliseconds = 200), line 275 in "sem.c"
[5] NamedSem_TimedWait(self = 0xff2d9ec4, milliseconds = 200), line 195 in "sem.h"
[6] TestSystem::Run::PipeThread(this = 0xff2da314), line 643 in "Run.cpp"
[7] _PipeThread(param = (nil)), line 632 in "Run.cpp"
[8] _Wrapper(param = 0x12a600), line 221 in "thread.c"
(dbx) where t@3
dbx: no thread with id 3
(dbx) where t@79
current thread: t@79
=>[1] __lwp_park(0x4, 0x0, 0x0, 0x1, 0x0, 0xfed90080), at 0xf3345478
[2] cond_sleep_queue(0x2f918, 0x2f928, 0x0, 0x0, 0x0, 0x0), at 0xf333f020
[3] cond_wait_queue(0x2f918, 0x2f928, 0x0, 0x0, 0x0, 0x0), at 0xf333f288
[4] cond_wait(0x2f918, 0x2f928, 0x0, 0x0, 0xfd5d1240, 0x0), at 0xf333f8d0
[5] pthread_cond_wait(0x2f918, 0x2f928, 0x0, 0x0, 0x2f928, 0xfffc00), at 0xf333f90c
[6] CondLock_Wait(key = 1161952, destination = 0x11bb18, comparand = 0, spinCount = 4294967295U), line 1199 in "condlockposix.c"
[7] InteractionProtocolHandler_ThreadShutdown(_application = 0x11bae0), line 2435 in "InteractionProtocolHandler.c"
[8] _Wrapper(param = 0x115a58), line 221 in "thread.c"

The compilers for hpux, AIX and solaris do not support mixed width literal concatenation. The macros defined in tests/pal/test_pal_strings.h depend on that feature. It was necessary to deactivate the test cases that uses those macros on those platforms, for wchar.

This shows up in regress