Eric
I reverted your last two commits, since the new thorn list didn't work for me. It created a file GetComponents in the main Cactus directory pointing to ../repos, which is the wrong path.
Checking out Cactus is a crucial step; once this is broken, everybody will have to update manually (since GetComponents doesn't work any more and will have been deleted!). Let's debug this, and let's have a few other people try things before proceeding.
-erik
On Fri, Sep 17, 2010 at 5:03 PM, eric9@mac.com wrote:
User: eric9 Date: 2010/09/17 05:03 PM
Modified: /trunk/ einsteintoolkit.th
Log: update to include new location of GetComponents
File Changes:
Directory: /trunk/
File [modified]: einsteintoolkit.th Delta lines: +11 -4 =================================================================== --- trunk/einsteintoolkit.th 2010-09-08 20:51:39 UTC (rev 37) +++ trunk/einsteintoolkit.th 2010-09-17 22:03:56 UTC (rev 38) @@ -111,7 +111,7 @@ !TARGET = $ARR !TYPE = svn !URL = https://svn.einsteintoolkit.org/cactus/$1/$2/trunk -!CHECKOUT = +!CHECKOUT = EinsteinAnalysis/ADMAnalysis EinsteinAnalysis/ADMConstraints EinsteinAnalysis/AHFinder @@ -190,7 +190,7 @@ !TYPE = svn !URL = http://svn.aei.mpg.de/numrel/$1/$2/trunk !AUTH_URL = https://svn.aei.mpg.de/numrel/$1/$2/trunk -!CHECKOUT = +!CHECKOUT = AEIThorns/AEILocalInterp
@@ -198,7 +198,7 @@ !TARGET = $ARR !TYPE = svn !URL = https://svn.cct.lsu.edu/repos/numrel/$1/$2/trunk -!CHECKOUT = +!CHECKOUT = LSUThorns/QuasiLocalMeasures LSUThorns/SummationByParts
@@ -210,7 +210,7 @@ !URL = git://github.com/ianhinder/Kranc.git !AUTH_URL = git@github.com:ianhinder/Kranc.git !REPO_PATH= Auxiliary/Cactus -!CHECKOUT = +!CHECKOUT = KrancNumericalTools/GenericFD
@@ -258,3 +258,10 @@ Carpet/CarpetSlab Carpet/CarpetTracker Carpet/LoopControl
+# GetComponents, to keep updating +!TARGET = $ROOT +!TYPE = git +!URL = git://github.com/gridaphobe/GetComponents.git +!AUTH_URL = git@github.com:gridaphobe/GetComponents.git +!CHECKOUT = GetComponents \ No newline at end of file
Commits mailing list Commits@einsteintoolkit.org http://lists.einsteintoolkit.org/mailman/listinfo/commits
Did you use the version from github? I was getting the exact same error with the svn trunk version, which was due to a bad assumption that !CHECKOUT would contain something of the form "xxx/yyy". What I could do is recommit the working version of GetComponents to svn for now, until there can be a more stable transition to git?
Eric
On Sep 17, 2010, at 6:52 PM, Erik Schnetter wrote:
Eric
I reverted your last two commits, since the new thorn list didn't work for me. It created a file GetComponents in the main Cactus directory pointing to ../repos, which is the wrong path.
Checking out Cactus is a crucial step; once this is broken, everybody will have to update manually (since GetComponents doesn't work any more and will have been deleted!). Let's debug this, and let's have a few other people try things before proceeding.
-erik
On Fri, Sep 17, 2010 at 5:03 PM, eric9@mac.com wrote:
User: eric9 Date: 2010/09/17 05:03 PM
Modified: /trunk/ einsteintoolkit.th
Log: update to include new location of GetComponents
File Changes:
Directory: /trunk/
File [modified]: einsteintoolkit.th Delta lines: +11 -4 =================================================================== --- trunk/einsteintoolkit.th 2010-09-08 20:51:39 UTC (rev 37) +++ trunk/einsteintoolkit.th 2010-09-17 22:03:56 UTC (rev 38) @@ -111,7 +111,7 @@ !TARGET = $ARR !TYPE = svn !URL = https://svn.einsteintoolkit.org/cactus/$1/$2/trunk -!CHECKOUT = +!CHECKOUT = EinsteinAnalysis/ADMAnalysis EinsteinAnalysis/ADMConstraints EinsteinAnalysis/AHFinder @@ -190,7 +190,7 @@ !TYPE = svn !URL = http://svn.aei.mpg.de/numrel/$1/$2/trunk !AUTH_URL = https://svn.aei.mpg.de/numrel/$1/$2/trunk -!CHECKOUT = +!CHECKOUT = AEIThorns/AEILocalInterp
@@ -198,7 +198,7 @@ !TARGET = $ARR !TYPE = svn !URL = https://svn.cct.lsu.edu/repos/numrel/$1/$2/trunk -!CHECKOUT = +!CHECKOUT = LSUThorns/QuasiLocalMeasures LSUThorns/SummationByParts
@@ -210,7 +210,7 @@ !URL = git://github.com/ianhinder/Kranc.git !AUTH_URL = git@github.com:ianhinder/Kranc.git !REPO_PATH= Auxiliary/Cactus -!CHECKOUT = +!CHECKOUT = KrancNumericalTools/GenericFD
@@ -258,3 +258,10 @@ Carpet/CarpetSlab Carpet/CarpetTracker Carpet/LoopControl
+# GetComponents, to keep updating +!TARGET = $ROOT +!TYPE = git +!URL = git://github.com/gridaphobe/GetComponents.git +!AUTH_URL = git@github.com:gridaphobe/GetComponents.git +!CHECKOUT = GetComponents \ No newline at end of file
Commits mailing list Commits@einsteintoolkit.org http://lists.einsteintoolkit.org/mailman/listinfo/commits
-- Erik Schnetter schnetter@cct.lsu.edu http://www.cct.lsu.edu/~eschnett/ _______________________________________________ Users mailing list Users@einsteintoolkit.org http://lists.einsteintoolkit.org/mailman/listinfo/users
----- Original message -----
Eric
I reverted your last two commits, since the new thorn list didn't work for me. It created a file GetComponents in the main Cactus directory pointing to ../repos, which is the wrong path.
Checking out Cactus is a crucial step; once this is broken, everybody will have to update manually (since GetComponents doesn't work any more and will have been deleted!). Let's debug this, and let's have a few other people try things before proceeding.
And a reminder: as I understood it, the policy is that all nontrivial commits to the trunk of the ET (excluding the MHD part of grhydro which is new code under heavy development) are to be discussed on the mailing list before being applied. (The policy might be too strict and tedious, but we should either try to stick to it or change it.) Part of the discussion would likely include other people testing the patches.
-erik
On Fri, Sep 17, 2010 at 5:03 PM, eric9@mac.com wrote:
User: eric9 Date: 2010/09/17 05:03 PM
Modified: /trunk/ einsteintoolkit.th
Log: update to include new location of GetComponents
File Changes:
Directory: /trunk/
File [modified]: einsteintoolkit.th Delta lines: +11 -4 =================================================================== --- trunk/einsteintoolkit.th 2010-09-08 20:51:39 UTC (rev 37) +++ trunk/einsteintoolkit.th 2010-09-17 22:03:56 UTC (rev 38) @@ -111,7 +111,7 @@ !TARGET = $ARR !TYPE = svn !URL = https://svn.einsteintoolkit.org/cactus/$1/$2/trunk -!CHECKOUT = +!CHECKOUT = EinsteinAnalysis/ADMAnalysis EinsteinAnalysis/ADMConstraints EinsteinAnalysis/AHFinder @@ -190,7 +190,7 @@ !TYPE = svn !URL = http://svn.aei.mpg.de/numrel/$1/$2/trunk !AUTH_URL = https://svn.aei.mpg.de/numrel/$1/$2/trunk -!CHECKOUT = +!CHECKOUT = AEIThorns/AEILocalInterp
@@ -198,7 +198,7 @@ !TARGET = $ARR !TYPE = svn !URL = https://svn.cct.lsu.edu/repos/numrel/$1/$2/trunk -!CHECKOUT = +!CHECKOUT = LSUThorns/QuasiLocalMeasures LSUThorns/SummationByParts
@@ -210,7 +210,7 @@ !URL = git://github.com/ianhinder/Kranc.git !AUTH_URL = git@github.com:ianhinder/Kranc.git !REPO_PATH= Auxiliary/Cactus -!CHECKOUT = +!CHECKOUT = KrancNumericalTools/GenericFD
@@ -258,3 +258,10 @@ Carpet/CarpetSlab Carpet/CarpetTracker Carpet/LoopControl
+# GetComponents, to keep updating +!TARGET = $ROOT +!TYPE = git +!URL = git://github.com/gridaphobe/GetComponents.git +!AUTH_URL = git@github.com:gridaphobe/GetComponents.git +!CHECKOUT = GetComponents \ No newline at end of file
Commits mailing list Commits@einsteintoolkit.org http://lists.einsteintoolkit.org/mailman/listinfo/commits
-- Erik Schnetter schnetter@cct.lsu.edu http://www.cct.lsu.edu/~eschnett/ _______________________________________________ Users mailing list Users@einsteintoolkit.org http://lists.einsteintoolkit.org/mailman/listinfo/users
Ok sorry about the mess there. I have created a patch that fixes the issue Erik was having with GetComponents, but I would like some people to test it since I get a segfault of all things when I issue a very specific command. The segfault occurs when i run
/Cactus/utils/Scripts/GetComponents -a -p --root=TEST Cactus/manifest/my_einsteintoolkit.th
my_einsteintoolkit.th is slightly modified to include the Flickr and Twitter thorns, as well as a few others; however it never reaches those thorns. The segfault always occurs immediately after GetComponents has been cloned and symlinked. The weird thing is that if I modify the command at all, it no longer occurs. If I add any level of verbosity, if I disable parallel checkout, if I use einsteintoolkit.th (which for me still includes GetComponents from github), or if I call GetComponents from the git repository (which is identical to my patched svn version) the entire checkout succeeds.
I'm attaching the patch and my_einsteintoolkit.th. Let me know if anyone else can reproduce this error, perhaps it's a problem with my installation of perl or git.
Eric
On Sep 18, 2010, at 3:48 AM, Ian Hinder wrote:
----- Original message -----
Eric
I reverted your last two commits, since the new thorn list didn't work for me. It created a file GetComponents in the main Cactus directory pointing to ../repos, which is the wrong path.
Checking out Cactus is a crucial step; once this is broken, everybody will have to update manually (since GetComponents doesn't work any more and will have been deleted!). Let's debug this, and let's have a few other people try things before proceeding.
And a reminder: as I understood it, the policy is that all nontrivial commits to the trunk of the ET (excluding the MHD part of grhydro which is new code under heavy development) are to be discussed on the mailing list before being applied. (The policy might be too strict and tedious, but we should either try to stick to it or change it.) Part of the discussion would likely include other people testing the patches.
-erik
On Fri, Sep 17, 2010 at 5:03 PM, eric9@mac.com wrote:
User: eric9 Date: 2010/09/17 05:03 PM
Modified: /trunk/ einsteintoolkit.th
Log: update to include new location of GetComponents
File Changes:
Directory: /trunk/
File [modified]: einsteintoolkit.th Delta lines: +11 -4 =================================================================== --- trunk/einsteintoolkit.th 2010-09-08 20:51:39 UTC (rev 37) +++ trunk/einsteintoolkit.th 2010-09-17 22:03:56 UTC (rev 38) @@ -111,7 +111,7 @@ !TARGET = $ARR !TYPE = svn !URL = https://svn.einsteintoolkit.org/cactus/$1/$2/trunk -!CHECKOUT = +!CHECKOUT = EinsteinAnalysis/ADMAnalysis EinsteinAnalysis/ADMConstraints EinsteinAnalysis/AHFinder @@ -190,7 +190,7 @@ !TYPE = svn !URL = http://svn.aei.mpg.de/numrel/$1/$2/trunk !AUTH_URL = https://svn.aei.mpg.de/numrel/$1/$2/trunk -!CHECKOUT = +!CHECKOUT = AEIThorns/AEILocalInterp
@@ -198,7 +198,7 @@ !TARGET = $ARR !TYPE = svn !URL = https://svn.cct.lsu.edu/repos/numrel/$1/$2/trunk -!CHECKOUT = +!CHECKOUT = LSUThorns/QuasiLocalMeasures LSUThorns/SummationByParts
@@ -210,7 +210,7 @@ !URL = git://github.com/ianhinder/Kranc.git !AUTH_URL = git@github.com:ianhinder/Kranc.git !REPO_PATH= Auxiliary/Cactus -!CHECKOUT = +!CHECKOUT = KrancNumericalTools/GenericFD
@@ -258,3 +258,10 @@ Carpet/CarpetSlab Carpet/CarpetTracker Carpet/LoopControl
+# GetComponents, to keep updating +!TARGET = $ROOT +!TYPE = git +!URL = git://github.com/gridaphobe/GetComponents.git +!AUTH_URL = git@github.com:gridaphobe/GetComponents.git +!CHECKOUT = GetComponents \ No newline at end of file
Commits mailing list Commits@einsteintoolkit.org http://lists.einsteintoolkit.org/mailman/listinfo/commits
-- Erik Schnetter schnetter@cct.lsu.edu http://www.cct.lsu.edu/~eschnett/ _______________________________________________ Users mailing list Users@einsteintoolkit.org http://lists.einsteintoolkit.org/mailman/listinfo/users
Users mailing list Users@einsteintoolkit.org http://lists.einsteintoolkit.org/mailman/listinfo/users
A segfault in a perl script indicates an error in the perl interpreter, or that you are running out of memory. Another possibility may be that you are overwriting the perl script that is currently executing -- I don't know what happens in this case, but I think nothing should happen since the perl interpreter reads the script file in the beginning and then ignores all changes.
Do you have another perl version with which you can try?
Since you mention that you run in parallel, maybe there is another resource (not memory) that is exhausted, such as threads or processes or stack space. Maybe using fewer workers would change things?
-erik
On Sat, Sep 18, 2010 at 10:42 AM, Eric Seidel eric@eseidel.org wrote:
Ok sorry about the mess there. I have created a patch that fixes the issue Erik was having with GetComponents, but I would like some people to test it since I get a segfault of all things when I issue a very specific command. The segfault occurs when i run /Cactus/utils/Scripts/GetComponents -a -p --root=TEST Cactus/manifest/my_einsteintoolkit.th my_einsteintoolkit.th is slightly modified to include the Flickr and Twitter thorns, as well as a few others; however it never reaches those thorns. The segfault always occurs immediately after GetComponents has been cloned and symlinked. The weird thing is that if I modify the command at all, it no longer occurs. If I add any level of verbosity, if I disable parallel checkout, if I use einsteintoolkit.th (which for me still includes GetComponents from github), or if I call GetComponents from the git repository (which is identical to my patched svn version) the entire checkout succeeds. I'm attaching the patch and my_einsteintoolkit.th. Let me know if anyone else can reproduce this error, perhaps it's a problem with my installation of perl or git. Eric
On Sep 18, 2010, at 3:48 AM, Ian Hinder wrote:
----- Original message -----
Eric
I reverted your last two commits, since the new thorn list didn't work for me. It created a file GetComponents in the main Cactus directory pointing to ../repos, which is the wrong path.
Checking out Cactus is a crucial step; once this is broken, everybody will have to update manually (since GetComponents doesn't work any more and will have been deleted!). Let's debug this, and let's have a few other people try things before proceeding.
And a reminder: as I understood it, the policy is that all nontrivial commits to the trunk of the ET (excluding the MHD part of grhydro which is new code under heavy development) are to be discussed on the mailing list before being applied. (The policy might be too strict and tedious, but we should either try to stick to it or change it.) Part of the discussion would likely include other people testing the patches.
-erik
On Fri, Sep 17, 2010 at 5:03 PM, eric9@mac.com wrote:
User: eric9 Date: 2010/09/17 05:03 PM
Modified: /trunk/ einsteintoolkit.th
Log: update to include new location of GetComponents
File Changes:
Directory: /trunk/
File [modified]: einsteintoolkit.th Delta lines: +11 -4 =================================================================== --- trunk/einsteintoolkit.th 2010-09-08 20:51:39 UTC (rev 37) +++ trunk/einsteintoolkit.th 2010-09-17 22:03:56 UTC (rev 38) @@ -111,7 +111,7 @@ !TARGET = $ARR !TYPE = svn !URL = https://svn.einsteintoolkit.org/cactus/$1/$2/trunk -!CHECKOUT = +!CHECKOUT = EinsteinAnalysis/ADMAnalysis EinsteinAnalysis/ADMConstraints EinsteinAnalysis/AHFinder @@ -190,7 +190,7 @@ !TYPE = svn !URL = http://svn.aei.mpg.de/numrel/$1/$2/trunk !AUTH_URL = https://svn.aei.mpg.de/numrel/$1/$2/trunk -!CHECKOUT = +!CHECKOUT = AEIThorns/AEILocalInterp
@@ -198,7 +198,7 @@ !TARGET = $ARR !TYPE = svn !URL = https://svn.cct.lsu.edu/repos/numrel/$1/$2/trunk -!CHECKOUT = +!CHECKOUT = LSUThorns/QuasiLocalMeasures LSUThorns/SummationByParts
@@ -210,7 +210,7 @@ !URL = git://github.com/ianhinder/Kranc.git !AUTH_URL = git@github.com:ianhinder/Kranc.git !REPO_PATH= Auxiliary/Cactus -!CHECKOUT = +!CHECKOUT = KrancNumericalTools/GenericFD
@@ -258,3 +258,10 @@ Carpet/CarpetSlab Carpet/CarpetTracker Carpet/LoopControl
+# GetComponents, to keep updating +!TARGET = $ROOT +!TYPE = git +!URL = git://github.com/gridaphobe/GetComponents.git +!AUTH_URL = git@github.com:gridaphobe/GetComponents.git +!CHECKOUT = GetComponents \ No newline at end of file
Commits mailing list Commits@einsteintoolkit.org http://lists.einsteintoolkit.org/mailman/listinfo/commits
-- Erik Schnetter schnetter@cct.lsu.edu http://www.cct.lsu.edu/~eschnett/ _______________________________________________ Users mailing list Users@einsteintoolkit.org http://lists.einsteintoolkit.org/mailman/listinfo/users
Users mailing list Users@einsteintoolkit.org http://lists.einsteintoolkit.org/mailman/listinfo/users
Users mailing list Users@einsteintoolkit.org http://lists.einsteintoolkit.org/mailman/listinfo/users
I just tried again after installing perl 5.12 and it worked. I guess maybe it was an issue with the OSX version of perl...
Have you tried checking out GetComponents from github with the patch I sent earlier? It should fix the symlink error.
Thanks,
Eric
On Sep 21, 2010, at 7:32 PM, Erik Schnetter wrote:
A segfault in a perl script indicates an error in the perl interpreter, or that you are running out of memory. Another possibility may be that you are overwriting the perl script that is currently executing -- I don't know what happens in this case, but I think nothing should happen since the perl interpreter reads the script file in the beginning and then ignores all changes.
Do you have another perl version with which you can try?
Since you mention that you run in parallel, maybe there is another resource (not memory) that is exhausted, such as threads or processes or stack space. Maybe using fewer workers would change things?
-erik
On Sat, Sep 18, 2010 at 10:42 AM, Eric Seidel eric@eseidel.org wrote:
Ok sorry about the mess there. I have created a patch that fixes the issue Erik was having with GetComponents, but I would like some people to test it since I get a segfault of all things when I issue a very specific command. The segfault occurs when i run /Cactus/utils/Scripts/GetComponents -a -p --root=TEST Cactus/manifest/my_einsteintoolkit.th my_einsteintoolkit.th is slightly modified to include the Flickr and Twitter thorns, as well as a few others; however it never reaches those thorns. The segfault always occurs immediately after GetComponents has been cloned and symlinked. The weird thing is that if I modify the command at all, it no longer occurs. If I add any level of verbosity, if I disable parallel checkout, if I use einsteintoolkit.th (which for me still includes GetComponents from github), or if I call GetComponents from the git repository (which is identical to my patched svn version) the entire checkout succeeds. I'm attaching the patch and my_einsteintoolkit.th. Let me know if anyone else can reproduce this error, perhaps it's a problem with my installation of perl or git. Eric
On Sep 18, 2010, at 3:48 AM, Ian Hinder wrote:
----- Original message -----
Eric
I reverted your last two commits, since the new thorn list didn't work for me. It created a file GetComponents in the main Cactus directory pointing to ../repos, which is the wrong path.
Checking out Cactus is a crucial step; once this is broken, everybody will have to update manually (since GetComponents doesn't work any more and will have been deleted!). Let's debug this, and let's have a few other people try things before proceeding.
And a reminder: as I understood it, the policy is that all nontrivial commits to the trunk of the ET (excluding the MHD part of grhydro which is new code under heavy development) are to be discussed on the mailing list before being applied. (The policy might be too strict and tedious, but we should either try to stick to it or change it.) Part of the discussion would likely include other people testing the patches.
-erik
On Fri, Sep 17, 2010 at 5:03 PM, eric9@mac.com wrote:
User: eric9 Date: 2010/09/17 05:03 PM
Modified: /trunk/ einsteintoolkit.th
Log: update to include new location of GetComponents
File Changes:
Directory: /trunk/
File [modified]: einsteintoolkit.th Delta lines: +11 -4 =================================================================== --- trunk/einsteintoolkit.th 2010-09-08 20:51:39 UTC (rev 37) +++ trunk/einsteintoolkit.th 2010-09-17 22:03:56 UTC (rev 38) @@ -111,7 +111,7 @@ !TARGET = $ARR !TYPE = svn !URL = https://svn.einsteintoolkit.org/cactus/$1/$2/trunk -!CHECKOUT = +!CHECKOUT = EinsteinAnalysis/ADMAnalysis EinsteinAnalysis/ADMConstraints EinsteinAnalysis/AHFinder @@ -190,7 +190,7 @@ !TYPE = svn !URL = http://svn.aei.mpg.de/numrel/$1/$2/trunk !AUTH_URL = https://svn.aei.mpg.de/numrel/$1/$2/trunk -!CHECKOUT = +!CHECKOUT = AEIThorns/AEILocalInterp
@@ -198,7 +198,7 @@ !TARGET = $ARR !TYPE = svn !URL = https://svn.cct.lsu.edu/repos/numrel/$1/$2/trunk -!CHECKOUT = +!CHECKOUT = LSUThorns/QuasiLocalMeasures LSUThorns/SummationByParts
@@ -210,7 +210,7 @@ !URL = git://github.com/ianhinder/Kranc.git !AUTH_URL = git@github.com:ianhinder/Kranc.git !REPO_PATH= Auxiliary/Cactus -!CHECKOUT = +!CHECKOUT = KrancNumericalTools/GenericFD
@@ -258,3 +258,10 @@ Carpet/CarpetSlab Carpet/CarpetTracker Carpet/LoopControl
+# GetComponents, to keep updating +!TARGET = $ROOT +!TYPE = git +!URL = git://github.com/gridaphobe/GetComponents.git +!AUTH_URL = git@github.com:gridaphobe/GetComponents.git +!CHECKOUT = GetComponents \ No newline at end of file
Commits mailing list Commits@einsteintoolkit.org http://lists.einsteintoolkit.org/mailman/listinfo/commits
-- Erik Schnetter schnetter@cct.lsu.edu http://www.cct.lsu.edu/~eschnett/ _______________________________________________ Users mailing list Users@einsteintoolkit.org http://lists.einsteintoolkit.org/mailman/listinfo/users
Users mailing list Users@einsteintoolkit.org http://lists.einsteintoolkit.org/mailman/listinfo/users
Users mailing list Users@einsteintoolkit.org http://lists.einsteintoolkit.org/mailman/listinfo/users
-- Erik Schnetter schnetter@cct.lsu.edu http://www.cct.lsu.edu/~eschnett/
Yes, the patch works. (However, the patch is a bit strange in that its indentation is off, and it contains comment calling the solution "a hack". Is there a better solution?)
-erik
On Tue, Sep 21, 2010 at 9:00 PM, Eric Seidel eric@eseidel.org wrote:
I just tried again after installing perl 5.12 and it worked. I guess maybe it was an issue with the OSX version of perl...
Have you tried checking out GetComponents from github with the patch I sent earlier? It should fix the symlink error.
Thanks,
Eric
On Sep 21, 2010, at 7:32 PM, Erik Schnetter wrote:
A segfault in a perl script indicates an error in the perl interpreter, or that you are running out of memory. Another possibility may be that you are overwriting the perl script that is currently executing -- I don't know what happens in this case, but I think nothing should happen since the perl interpreter reads the script file in the beginning and then ignores all changes.
Do you have another perl version with which you can try?
Since you mention that you run in parallel, maybe there is another resource (not memory) that is exhausted, such as threads or processes or stack space. Maybe using fewer workers would change things?
-erik
On Sat, Sep 18, 2010 at 10:42 AM, Eric Seidel eric@eseidel.org wrote:
Ok sorry about the mess there. I have created a patch that fixes the issue Erik was having with GetComponents, but I would like some people to test it since I get a segfault of all things when I issue a very specific command. The segfault occurs when i run /Cactus/utils/Scripts/GetComponents -a -p --root=TEST Cactus/manifest/my_einsteintoolkit.th my_einsteintoolkit.th is slightly modified to include the Flickr and Twitter thorns, as well as a few others; however it never reaches those thorns. The segfault always occurs immediately after GetComponents has been cloned and symlinked. The weird thing is that if I modify the command at all, it no longer occurs. If I add any level of verbosity, if I disable parallel checkout, if I use einsteintoolkit.th (which for me still includes GetComponents from github), or if I call GetComponents from the git repository (which is identical to my patched svn version) the entire checkout succeeds. I'm attaching the patch and my_einsteintoolkit.th. Let me know if anyone else can reproduce this error, perhaps it's a problem with my installation of perl or git. Eric
On Sep 18, 2010, at 3:48 AM, Ian Hinder wrote:
----- Original message -----
Eric
I reverted your last two commits, since the new thorn list didn't work for me. It created a file GetComponents in the main Cactus directory pointing to ../repos, which is the wrong path.
Checking out Cactus is a crucial step; once this is broken, everybody will have to update manually (since GetComponents doesn't work any more and will have been deleted!). Let's debug this, and let's have a few other people try things before proceeding.
And a reminder: as I understood it, the policy is that all nontrivial commits to the trunk of the ET (excluding the MHD part of grhydro which is new code under heavy development) are to be discussed on the mailing list before being applied. (The policy might be too strict and tedious, but we should either try to stick to it or change it.) Part of the discussion would likely include other people testing the patches.
-erik
On Fri, Sep 17, 2010 at 5:03 PM, eric9@mac.com wrote:
User: eric9 Date: 2010/09/17 05:03 PM
Modified: /trunk/ einsteintoolkit.th
Log: update to include new location of GetComponents
File Changes:
Directory: /trunk/
File [modified]: einsteintoolkit.th Delta lines: +11 -4 =================================================================== --- trunk/einsteintoolkit.th 2010-09-08 20:51:39 UTC (rev 37) +++ trunk/einsteintoolkit.th 2010-09-17 22:03:56 UTC (rev 38) @@ -111,7 +111,7 @@ !TARGET = $ARR !TYPE = svn !URL = https://svn.einsteintoolkit.org/cactus/$1/$2/trunk -!CHECKOUT = +!CHECKOUT = EinsteinAnalysis/ADMAnalysis EinsteinAnalysis/ADMConstraints EinsteinAnalysis/AHFinder @@ -190,7 +190,7 @@ !TYPE = svn !URL = http://svn.aei.mpg.de/numrel/$1/$2/trunk !AUTH_URL = https://svn.aei.mpg.de/numrel/$1/$2/trunk -!CHECKOUT = +!CHECKOUT = AEIThorns/AEILocalInterp
@@ -198,7 +198,7 @@ !TARGET = $ARR !TYPE = svn !URL = https://svn.cct.lsu.edu/repos/numrel/$1/$2/trunk -!CHECKOUT = +!CHECKOUT = LSUThorns/QuasiLocalMeasures LSUThorns/SummationByParts
@@ -210,7 +210,7 @@ !URL = git://github.com/ianhinder/Kranc.git !AUTH_URL = git@github.com:ianhinder/Kranc.git !REPO_PATH= Auxiliary/Cactus -!CHECKOUT = +!CHECKOUT = KrancNumericalTools/GenericFD
@@ -258,3 +258,10 @@ Carpet/CarpetSlab Carpet/CarpetTracker Carpet/LoopControl
+# GetComponents, to keep updating +!TARGET = $ROOT +!TYPE = git +!URL = git://github.com/gridaphobe/GetComponents.git +!AUTH_URL = git@github.com:gridaphobe/GetComponents.git +!CHECKOUT = GetComponents \ No newline at end of file
Commits mailing list Commits@einsteintoolkit.org http://lists.einsteintoolkit.org/mailman/listinfo/commits
-- Erik Schnetter schnetter@cct.lsu.edu http://www.cct.lsu.edu/~eschnett/ _______________________________________________ Users mailing list Users@einsteintoolkit.org http://lists.einsteintoolkit.org/mailman/listinfo/users
Users mailing list Users@einsteintoolkit.org http://lists.einsteintoolkit.org/mailman/listinfo/users
Users mailing list Users@einsteintoolkit.org http://lists.einsteintoolkit.org/mailman/listinfo/users
-- Erik Schnetter schnetter@cct.lsu.edu http://www.cct.lsu.edu/~eschnett/
Users mailing list Users@einsteintoolkit.org http://lists.einsteintoolkit.org/mailman/listinfo/users
Somehow some tabs got in the code, I normally have BBEdit auto expand the tabs to be consistent.. I think the solution in the patch is kind of ugly from a design standpoint, which is probably why i called it a hack in the comment, but it is a proper solution. The other way I can see to address it would be to check for a path separator in $checkout, and then conditionally split. But then I would need to see if $checkout_dir is defined before creating the relative path, so it ends up being the same amount or more code.
I don't think it matters much which way the fix is implemented, although I should probably modify the comment calling it a hack.
Eric
On Sep 22, 2010, at 10:52 PM, Erik Schnetter wrote:
Yes, the patch works. (However, the patch is a bit strange in that its indentation is off, and it contains comment calling the solution "a hack". Is there a better solution?)
-erik
On Tue, Sep 21, 2010 at 9:00 PM, Eric Seidel eric@eseidel.org wrote:
I just tried again after installing perl 5.12 and it worked. I guess maybe it was an issue with the OSX version of perl...
Have you tried checking out GetComponents from github with the patch I sent earlier? It should fix the symlink error.
Thanks,
Eric
On Sep 21, 2010, at 7:32 PM, Erik Schnetter wrote:
A segfault in a perl script indicates an error in the perl interpreter, or that you are running out of memory. Another possibility may be that you are overwriting the perl script that is currently executing -- I don't know what happens in this case, but I think nothing should happen since the perl interpreter reads the script file in the beginning and then ignores all changes.
Do you have another perl version with which you can try?
Since you mention that you run in parallel, maybe there is another resource (not memory) that is exhausted, such as threads or processes or stack space. Maybe using fewer workers would change things?
-erik
On Sat, Sep 18, 2010 at 10:42 AM, Eric Seidel eric@eseidel.org wrote:
Ok sorry about the mess there. I have created a patch that fixes the issue Erik was having with GetComponents, but I would like some people to test it since I get a segfault of all things when I issue a very specific command. The segfault occurs when i run /Cactus/utils/Scripts/GetComponents -a -p --root=TEST Cactus/manifest/my_einsteintoolkit.th my_einsteintoolkit.th is slightly modified to include the Flickr and Twitter thorns, as well as a few others; however it never reaches those thorns. The segfault always occurs immediately after GetComponents has been cloned and symlinked. The weird thing is that if I modify the command at all, it no longer occurs. If I add any level of verbosity, if I disable parallel checkout, if I use einsteintoolkit.th (which for me still includes GetComponents from github), or if I call GetComponents from the git repository (which is identical to my patched svn version) the entire checkout succeeds. I'm attaching the patch and my_einsteintoolkit.th. Let me know if anyone else can reproduce this error, perhaps it's a problem with my installation of perl or git. Eric
On Sep 18, 2010, at 3:48 AM, Ian Hinder wrote:
----- Original message -----
Eric
I reverted your last two commits, since the new thorn list didn't work for me. It created a file GetComponents in the main Cactus directory pointing to ../repos, which is the wrong path.
Checking out Cactus is a crucial step; once this is broken, everybody will have to update manually (since GetComponents doesn't work any more and will have been deleted!). Let's debug this, and let's have a few other people try things before proceeding.
And a reminder: as I understood it, the policy is that all nontrivial commits to the trunk of the ET (excluding the MHD part of grhydro which is new code under heavy development) are to be discussed on the mailing list before being applied. (The policy might be too strict and tedious, but we should either try to stick to it or change it.) Part of the discussion would likely include other people testing the patches.
-erik
On Fri, Sep 17, 2010 at 5:03 PM, eric9@mac.com wrote:
User: eric9 Date: 2010/09/17 05:03 PM
Modified: /trunk/ einsteintoolkit.th
Log: update to include new location of GetComponents
File Changes:
Directory: /trunk/
File [modified]: einsteintoolkit.th Delta lines: +11 -4 =================================================================== --- trunk/einsteintoolkit.th 2010-09-08 20:51:39 UTC (rev 37) +++ trunk/einsteintoolkit.th 2010-09-17 22:03:56 UTC (rev 38) @@ -111,7 +111,7 @@ !TARGET = $ARR !TYPE = svn !URL = https://svn.einsteintoolkit.org/cactus/$1/$2/trunk -!CHECKOUT = +!CHECKOUT = EinsteinAnalysis/ADMAnalysis EinsteinAnalysis/ADMConstraints EinsteinAnalysis/AHFinder @@ -190,7 +190,7 @@ !TYPE = svn !URL = http://svn.aei.mpg.de/numrel/$1/$2/trunk !AUTH_URL = https://svn.aei.mpg.de/numrel/$1/$2/trunk -!CHECKOUT = +!CHECKOUT = AEIThorns/AEILocalInterp
@@ -198,7 +198,7 @@ !TARGET = $ARR !TYPE = svn !URL = https://svn.cct.lsu.edu/repos/numrel/$1/$2/trunk -!CHECKOUT = +!CHECKOUT = LSUThorns/QuasiLocalMeasures LSUThorns/SummationByParts
@@ -210,7 +210,7 @@ !URL = git://github.com/ianhinder/Kranc.git !AUTH_URL = git@github.com:ianhinder/Kranc.git !REPO_PATH= Auxiliary/Cactus -!CHECKOUT = +!CHECKOUT = KrancNumericalTools/GenericFD
@@ -258,3 +258,10 @@ Carpet/CarpetSlab Carpet/CarpetTracker Carpet/LoopControl
+# GetComponents, to keep updating +!TARGET = $ROOT +!TYPE = git +!URL = git://github.com/gridaphobe/GetComponents.git +!AUTH_URL = git@github.com:gridaphobe/GetComponents.git +!CHECKOUT = GetComponents \ No newline at end of file
Commits mailing list Commits@einsteintoolkit.org http://lists.einsteintoolkit.org/mailman/listinfo/commits
-- Erik Schnetter schnetter@cct.lsu.edu http://www.cct.lsu.edu/~eschnett/ _______________________________________________ Users mailing list Users@einsteintoolkit.org http://lists.einsteintoolkit.org/mailman/listinfo/users
Users mailing list Users@einsteintoolkit.org http://lists.einsteintoolkit.org/mailman/listinfo/users
Users mailing list Users@einsteintoolkit.org http://lists.einsteintoolkit.org/mailman/listinfo/users
-- Erik Schnetter schnetter@cct.lsu.edu http://www.cct.lsu.edu/~eschnett/
Users mailing list Users@einsteintoolkit.org http://lists.einsteintoolkit.org/mailman/listinfo/users
-- Erik Schnetter schnetter@cct.lsu.edu http://www.cct.lsu.edu/~eschnett/
Ok I revised the comments to be more explanatory, and I ran the entire file through a tab to space filter to make sure the formatting looks right again :)
The new patch is a bit bigger because of the formatting, but there are no changes to the actual code beyond the original patch. I would like to apply it to the trunk unless there are any questions or objections.
In addition after applying this patch to GetComponents I would like to add the git version back to einsteintoolkit.th. Then after a certain period of time, maybe a month, we can be certain that everyone using the trunk will have the git version, and GetComponents can complete its transition to GitHub, removing the svn version from Utilities/Scripts.
Eric
Somehow some tabs got in the code, I normally have BBEdit auto expand the tabs to be consistent.. I think the solution in the patch is kind of ugly from a design standpoint, which is probably why i called it a hack in the comment, but it is a proper solution. The other way I can see to address it would be to check for a path separator in $checkout, and then conditionally split. But then I would need to see if $checkout_dir is defined before creating the relative path, so it ends up being the same amount or more code.
I don't think it matters much which way the fix is implemented, although I should probably modify the comment calling it a hack.
Eric
Please apply the patch to the svn version.
Many people update only if they encounter a problem. They don't update regularly, because this avoids pulling in new bugs. Therefore waiting a month won't do the trick. We'll have to think of something else, e.g. an email announcement or somesuch.
-erik
On Mon, Sep 27, 2010 at 9:48 PM, Eric Seidel eric@eseidel.org wrote:
Ok I revised the comments to be more explanatory, and I ran the entire file through a tab to space filter to make sure the formatting looks right again :)
The new patch is a bit bigger because of the formatting, but there are no changes to the actual code beyond the original patch. I would like to apply it to the trunk unless there are any questions or objections.
In addition after applying this patch to GetComponents I would like to add the git version back to einsteintoolkit.th. Then after a certain period of time, maybe a month, we can be certain that everyone using the trunk will have the git version, and GetComponents can complete its transition to GitHub, removing the svn version from Utilities/Scripts.
Eric
Somehow some tabs got in the code, I normally have BBEdit auto expand the tabs to be consistent.. I think the solution in the patch is kind of ugly from a design standpoint, which is probably why i called it a hack in the comment, but it is a proper solution. The other way I can see to address it would be to check for a path separator in $checkout, and then conditionally split. But then I would need to see if $checkout_dir is defined before creating the relative path, so it ends up being the same amount or more code.
I don't think it matters much which way the fix is implemented, although I should probably modify the comment calling it a hack.
users@lists.einsteintoolkit.org
-
Eric Seidel -
Erik Schnetter -
Ian Hinder