|
Notes |
|
|
(0012904)
|
|
Bill Hoffman
|
|
2008-07-31 13:38
|
|
|
Would you be willing to run a dashboard if we got cmake building? |
|
|
|
(0012907)
|
|
Scott McCreary
|
|
2008-07-31 14:01
|
|
|
Maybe. I do daily builds of Haiku on my PC. There may be a way that I can add a dashboard into that script. Contact me at scottmc2 at gmail com for more info. |
|
|
|
(0013129)
|
|
Scott McCreary
|
|
2008-08-21 13:31
|
|
|
|
|
(0013133)
|
|
Bill Hoffman
|
|
2008-08-21 14:56
|
|
|
What are the errors you have right now? |
|
|
|
(0013134)
|
|
Scott McCreary
|
|
2008-08-21 15:04
|
|
During make, it either errors out with this, or stalls just on the "loading initial cache file" part. This output is from 2.6.0 but I get similar results on 2.6.1 as well. If I copy and paste the g++ line it compiles that part ok, so it seems it's something just after that, that is causing the error/hangup. Something in the cmcurl code?
g++ -lroot -lbe -I/boot/home/cmake-2.6.0-haiku-test/Source -I/boot/home/cmake
-2.6.0-haiku-test/Bootstrap.cmk cmake.o cmakemain.o cmakewizard.o cmCommandArgu
mentLexer.o cmCommandArgumentParser.o cmCommandArgumentParserHelper.o cmDepends.
o cmDependsC.o cmDocumentationFormatter.o cmDocumentationFormatterText.o cmPolic
ies.o cmProperty.o cmPropertyMap.o cmPropertyDefinition.o cmPropertyDefinitionMa
p.o cmMakeDepend.o cmMakefile.o cmExportFileGenerator.o cmExportInstallFileGener
ator.o cmInstallDirectoryGenerator.o cmGeneratedFileStream.o cmGlobalGenerator.o
cmLocalGenerator.o cmInstallGenerator.o cmInstallExportGenerator.o cmInstallFil
esGenerator.o cmInstallScriptGenerator.o cmInstallTargetGenerator.o cmSourceFile
.o cmSourceFileLocation.o cmSystemTools.o cmVersion.o cmFileTimeComparison.o cmG
lobalUnixMakefileGenerator3.o cmLocalUnixMakefileGenerator3.o cmMakefileExecutab
leTargetGenerator.o cmMakefileLibraryTargetGenerator.o cmMakefileTargetGenerator
.o cmMakefileUtilityTargetGenerator.o cmBootstrapCommands.o cmCommands.o cmTarge
t.o cmTest.o cmCustomCommand.o cmDocumentVariables.o cmCacheManager.o cmListFile
Cache.o cmComputeLinkDepends.o cmComputeLinkInformation.o cmOrderDirectories.o c
mComputeTargetDepends.o cmComputeComponentGraph.o cmExprLexer.o cmExprParser.o c
mExprParserHelper.o cmListFileLexer.o Directory.o Glob.o RegularExpression.o Sys
temTools.o ProcessUNIX.o String.o System.o -o cmake
loading initial cache file /boot/home/cmake-2.6.0-haiku-test/Bootstrap.cmk/Initi
alCacheFlags.cmake
CMake Error at Utilities/cmcurl/CMake/OtherTests.cmake:89 (MESSAGE):
Unable to link function recv
Call Stack (most recent call first):
Utilities/cmcurl/CMakeLists.txt:689 (INCLUDE)
-- Configuring done
---------------------------------------------
Error when bootstrapping CMake:
Problem while running initial CMake
--------------------------------------------- |
|
|
|
(0013151)
|
|
Bill Hoffman
|
|
2008-08-21 22:46
|
|
That means it bootstraps fine, but when the bootstrap cmake is run it gets a cmake error here:
Utilities/cmcurl/CMake/OtherTests.cmake:89
Which means it can not find a recv function. Is there one on the system? If so can you create a small c program that uses recv that works? |
|
|
|
(0013386)
|
|
Bill Hoffman
|
|
2008-09-09 14:14
|
|
|
|
|
(0013390)
|
|
Scott McCreary
|
|
2008-09-09 14:47
|
|
I posted to the Haiku dev mailing list and got this back:
> recv() is the syscall for receiving data on a network socket. In order
> to use it you'll need a) a socket file descriptor, b) sys/socket.h,
> and c) to link to libnetwork.so.
So perhaps it's not including sys/socket.h or linking with -lnetwork?
I may have already tried doing that, but it's been a couple weeks since I've done anything with this. |
|
|
|
(0013393)
|
|
Bill Hoffman
|
|
2008-09-09 15:12
|
|
Sounds like all we need to do is add -lnetwork to get it working.
Try adding this:
Utilities/cmcurl/CMakeLists.txt
CHECK_LIBRARY_EXISTS_CONCAT("network" recv HAVE_LIBNETWORK) |
|
|
|
(0013401)
|
|
Scott McCreary
|
|
2008-09-10 16:48
|
|
|
|
|
(0013403)
|
|
Bill Hoffman
|
|
2008-09-10 16:59
|
|
It is stuck running the bootstrap cmake which is good. You should be able to kill that and run it by hand.
cd build-tree
./Bootstrap.cmk/cmake --debug-output --trace ../CMake |
|
|
|
(0013404)
|
|
Scott McCreary
|
|
2008-09-10 17:22
|
|
Nope.
That stalls as well. It prints out:
Running with debug output on.
And updates the CMakeSystem.cmake and CMakeOutput.log file like the previous run did, but then nothing...
Waited over 10 minutes and nothing seems to have happened.
Any other ideas to try? |
|
|
|
(0013407)
|
|
Bill Hoffman
|
|
2008-09-11 08:44
|
|
Is there a debugger ? I would try that if not you will have to add some print statements. Sounds like it is not getting very far.
You could add come std::cerr << "here " type of stuff into cmakemain.cxx also in cmake.cxx cmake::Run. Is there an strace on the system? |
|
|
|
(0013408)
|
|
Bill Hoffman
|
|
2008-09-11 08:45
|
|
|
Can I have remote access to the system? |
|
|
|
(0013437)
|
|
Scott McCreary
|
|
2008-09-11 14:05
|
|
|
|
|
(0013445)
|
|
Bill Hoffman
|
|
2008-09-11 15:22
|
|
OK, I got the vm machine working. The trouble seems to be in the process execution code. If you run
strace ./Bootstrap.cmk/cmake
It gets stuck after a fork. This is when cmake is testing the compiler. |
|
|
|
(0013446)
|
|
Scott McCreary
|
|
2008-09-12 01:05
|
|
|
|
|
(0013450)
|
|
Bill Hoffman
|
|
2008-09-12 11:17
|
|
|
I can not get past the bootstrap part, and am not sure which ifndef you removed to fix the problem. I have the vm machine running, and should be able to fix it. Are you sure that diff is current? |
|
|
|
(0013453)
|
|
Bill Hoffman
|
|
2008-09-12 11:25
|
|
|
Never mind, I have it working now, I will try to figure out the next part. Looks like it is very close. |
|
|
|
(0013456)
|
|
Bill Hoffman
|
|
2008-09-12 13:10
|
|
OK, I found one thing.
Modules/CMakeFortranCompilerId.F90.in
! is not a comment character in c or c++, you have several of these:
! Haiku also defines __BEOS__ so we must
! put it prior to the check for __BEOS__ |
|
|
|
(0013459)
|
|
Bill Hoffman
|
|
2008-09-12 15:30
|
|
OK, I found the platform issue. For some reason the popen code is getting duplicate output with a newline in it. The code is in Source/cmSystemTools.cxx bool RunCommandViaPopen. When uname -s is run with this code, it gets an output of
Haiku
Haiku
including the newline in the middle.
Any ideas? |
|
|
|
(0013460)
|
|
Bill Hoffman
|
|
2008-09-12 15:38
|
|
Never mind again, I found a work around...
while(!feof(cpipe))
{
if(verbose)
{
cmSystemTools::Stdout(buffer);
}
output += buffer;
buffer[0] = 0; // needed to add this line
fgets(buffer, BUFFER_SIZE, cpipe);
}
Now it finds the Platform file. |
|
|
|
(0013461)
|
|
Scott McCreary
|
|
2008-09-12 15:44
|
|
In Source/cmSystemTools.cxx there's three places where there is am ifdef __BEOS__, these "might" need to be changed to:
if defined __BEOS__ && !defined __HAIKU__
since Haiku is also defining __BEOS__ at least until they hit R2.
This may or may not help. |
|
|
|
(0013462)
|
|
Bill Hoffman
|
|
2008-09-12 15:52
|
|
How do I install cvs on this vm machine?
It is running the bootstrap right now. Looks like it might work. The popen thing was causing the last bit of trouble. I would like to check out cvs cmake on this so I can check in the code once I make the changes. |
|
|
|
(0013463)
|
|
Scott McCreary
|
|
2008-09-12 15:57
|
|
|
|
|
(0013640)
|
|
Bill Hoffman
|
|
2008-10-01 10:20
|
|
OK, this is working in CVS, I will merge the fixes into the 2.6 branch. Now we need to setup a nightly build for this so it can stay working.
You should be able to run a script like this one:
http://www.cdash.org/CDash/viewNotes.php?buildid=185566 [^]
ctest -S nameofscript.cmake -VV
That should create a nightly dashboard for Haiku. |
|
cmake-build-on-haiku-stalls.png (163,373) 2008-09-10 16:44