Dear All,
I've finished building binaries for the GCC frontends and am now testing
the 1.5 release branch on i386/Linux, Sparc/Solaris, and PowerPC/MacOS X.
I'm looking for volunteers to test LLVM 1.5 on platforms that we don't
have in house. I'm specifically looking for people who can do:
Itanium
FreeBSD/i386
Windows (Cygwin, MingW, etc)
I'm mainly interested in major problems with these platforms.
If you can help, please email the list with what platform you'll be testing.
To get a copy of what I'm testing, you need to check out a copy of the
release branch as follows (adding whatever options are needed to set
CVSROOT properly):
cvs co -r release_15 llvm
cvs co -r release_15 llvm-gcc
cd llvm/projects
cvs co -r release_15 llvm-test
If you have commit access, please don't commit anything to the release
branch without asking the list first.
Thanks in advance.
-- John T.
Alexander Friedman wrote:
I think the nightly test script can be used for this specific task,
after all, that's what it's designed to do! If an option were added to
it to test a particular branch, it could be used directly without any
other modifications. Am I correct?
Misha Brukman wrote:
Personally, I never use "gmake check."
What I do is checkout everything out, build it, run the tests in
"llvm/test," and then run all of the tests in the llvm-test CVS
module.
I think the nightly test script can be used for this specific task,
after all, that's what it's designed to do! If an option were added to
it to test a particular branch, it could be used directly without any
other modifications. Am I correct?
For the most part. There are actually a couple of things that make the process different.
First, the nightly tester is designed to test LLVM given that you have a GCC frontend (CFE) already compiled. Part of my process is building CFE binaries and then testing them along with LLVM. The nightly tester does not build it's own CFE; it would have to be adjusted to do this.
Second, everyone is testing the CVS mainline. Having something that tests a specific CVS label would be good. That's a small mod to the script, as you suggested above.
-- John T.
Note that both of these things could be optional changes to the nightly test script and have been requested in the past. In particular, though the first one would require a lot more build time, it would be quite useful for some people.
If anyone has the perl skills and the inclination to hack on the nightly tester script, it would be a cool enhancement...
-Chris
I haven't specifically tried the 1.5 branch yet on FreeBSD, but the current version just before you branched it passes gmake check and fails four llvm-test tests: siod, SPASS, archie, and plot2fig. SPASS needs porting work to compile correctly on FreeBSD. I haven't analyzed the other three yet, but they weren't failing two days earlier.
John Criswell wrote:
I haven't specifically tried the 1.5 branch yet on FreeBSD, but the current version just before you branched it passes gmake check and fails four llvm-test tests: siod, SPASS, archie, and plot2fig. SPASS needs porting work to compile correctly on FreeBSD. I haven't analyzed the other three yet, but they weren't failing two days earlier.
Chances are they weren't even showing up on the nightly tests before. I recently added a patch that shows test which don't build with the native compiler.
In other words, if this is the case, nothing to see here. Move along, move along...
-Chris
John Criswell wrote:
Dear All,
I've finished building binaries for the GCC frontends and am now testing
the 1.5 release branch on i386/Linux, Sparc/Solaris, and PowerPC/MacOS X.
I'm looking for volunteers to test LLVM 1.5 on platforms that we don't
have in house. I'm specifically looking for people who can do:
Itanium
FreeBSD/i386
Windows (Cygwin, MingW, etc)
I'm mainly interested in major problems with these platforms.
If you can help, please email the list with what platform you'll be testing.
To get a copy of what I'm testing, you need to check out a copy of the
release branch as follows (adding whatever options are needed to set
CVSROOT properly):
cvs co -r release_15 llvm
cvs co -r release_15 llvm-gcc
cd llvm/projects
cvs co -r release_15 llvm-test
If you have commit access, please don't commit anything to the release
branch without asking the list first.
Thanks in advance.
-- John T.
_______________________________________________
LLVM Developers mailing list
LLVMdev@cs.uiuc.edu http://llvm.cs.uiuc.edu
http://mail.cs.uiuc.edu/mailman/listinfo/llvmdev
-Chris
siod fails with gcc link errors: cannot find -ldl
archie and plot2fig fail with gcc link errors: re_comp and re_exec are undefined
Chris Lattner wrote:
siod fails with gcc link errors: cannot find -ldl
archie and plot2fig fail with gcc link errors: re_comp and re_exec are undefined
Again, the reason that these are showing up is because I changed the makefiles. They were failing before, just not notifying us. If the native compiler fails, we have nothing to compare against.
As such (again, if this is the case), these are not regressions. However, getting them work on FreeBSD would be a good thing to do for LLVM CVS.
-Chris
Chris Lattner wrote:
I haven't specifically tried the 1.5 branch yet on FreeBSD, but the current version just before you branched it passes gmake check and fails four llvm-test tests: siod, SPASS, archie, and plot2fig. SPASS needs porting work to compile correctly on FreeBSD. I haven't analyzed the other three yet, but they weren't failing two days earlier.
Chances are they weren't even showing up on the nightly tests before. I recently added a patch that shows test which don't build with the native compiler.
In other words, if this is the case, nothing to see here. Move along, move along...
-Chris
John Criswell wrote:
Dear All,
I've finished building binaries for the GCC frontends and am now testing
the 1.5 release branch on i386/Linux, Sparc/Solaris, and PowerPC/MacOS X.
I'm looking for volunteers to test LLVM 1.5 on platforms that we don't
have in house. I'm specifically looking for people who can do:
Itanium
FreeBSD/i386
Windows (Cygwin, MingW, etc)
I'm mainly interested in major problems with these platforms.
If you can help, please email the list with what platform you'll be testing.
To get a copy of what I'm testing, you need to check out a copy of the
release branch as follows (adding whatever options are needed to set
CVSROOT properly):
cvs co -r release_15 llvm
cvs co -r release_15 llvm-gcc
cd llvm/projects
cvs co -r release_15 llvm-test
If you have commit access, please don't commit anything to the release
branch without asking the list first.
Thanks in advance.
-- John T.
_______________________________________________
LLVM Developers mailing list
LLVMdev@cs.uiuc.edu http://llvm.cs.uiuc.edu
http://mail.cs.uiuc.edu/mailman/listinfo/llvmdev
-Chris
_______________________________________________
LLVM Developers mailing list
LLVMdev@cs.uiuc.edu http://llvm.cs.uiuc.edu
http://mail.cs.uiuc.edu/mailman/listinfo/llvmdev
-Chris
No, these are probably not regressions. Anyway, building release_15 now.
Chris Lattner wrote:
Itanium
Everything checks out as expected. 
Duraid
> Itanium
Everything checks out as expected.
D*mn. You got to it before I did :).
Cool.