Hi.
I noticed that the thorn EinsteinInitialData/Hydro_InitExcision has a "par" subdirectory containing both parfiles *and* the resulting output directories; it looks just like a second set of test suites, except presumably they'd never be picked up by the testsuite process. Any reason why these are there (still true with new 2015_11 release)?
Thanks,
Bernard
On Fri, Nov 20, 2015 at 11:46:29AM -0500, Bernard Kelly wrote:
I noticed that the thorn EinsteinInitialData/Hydro_InitExcision has a "par" subdirectory containing both parfiles *and* the resulting output directories; it looks just like a second set of test suites, except presumably they'd never be picked up by the testsuite process. Any reason why these are there (still true with new 2015_11 release)?
Without looking into it deeply I seem to remember that the problem with those was that they were taking too long, and/or using too much memory to be used as regular test suite.
Frank
Thanks, Frank.
Wouldn't this be an argument for *removing* the data, then?
Or perhaps the test suite system could be equipped with a "short/medium/long" flag if people really wanted to avoid the most time-/memory-consuming tests?
This sounds like the kind of thing that might have been discussed before; I'm just objecting to anything other than .par files being in a "par" directory.
Bernard
On 20 November 2015 at 11:48, Frank Loeffler knarf@cct.lsu.edu wrote:
On Fri, Nov 20, 2015 at 11:46:29AM -0500, Bernard Kelly wrote:
I noticed that the thorn EinsteinInitialData/Hydro_InitExcision has a "par" subdirectory containing both parfiles *and* the resulting output directories; it looks just like a second set of test suites, except presumably they'd never be picked up by the testsuite process. Any reason why these are there (still true with new 2015_11 release)?
Without looking into it deeply I seem to remember that the problem with those was that they were taking too long, and/or using too much memory to be used as regular test suite.
Frank
On Fri, Nov 20, 2015 at 11:56:01AM -0500, Bernard Kelly wrote:
Wouldn't this be an argument for *removing* the data, then?
Or perhaps the test suite system could be equipped with a "short/medium/long" flag if people really wanted to avoid the most time-/memory-consuming tests?
This would be a good idea, and has been around for quite some time. It never made the threshold of being interesting enough for someone to implement though. So far.
Frank
These tests should probably be moved into a thorn of its own that is disabled by default. This makes it easy to run them if you want to (just compile the thorn), and otherwise they are out of the way. No matter what, the situation needs to be documented in a README, either in the new thorn, or a README file in the par directory.
Bernard -- do you want to submit a pull request on Bitbucket?
-erik
On Fri, Nov 20, 2015 at 10:56 AM, Bernard Kelly physicsbeany@gmail.com wrote:
Thanks, Frank.
Wouldn't this be an argument for *removing* the data, then?
Or perhaps the test suite system could be equipped with a "short/medium/long" flag if people really wanted to avoid the most time-/memory-consuming tests?
This sounds like the kind of thing that might have been discussed before; I'm just objecting to anything other than .par files being in a "par" directory.
Bernard
On 20 November 2015 at 11:48, Frank Loeffler knarf@cct.lsu.edu wrote:
On Fri, Nov 20, 2015 at 11:46:29AM -0500, Bernard Kelly wrote:
I noticed that the thorn EinsteinInitialData/Hydro_InitExcision has a "par" subdirectory containing both parfiles *and* the resulting output directories; it looks just like a second set of test suites, except presumably they'd never be picked up by the testsuite process. Any reason why these are there (still true with new 2015_11 release)?
Without looking into it deeply I seem to remember that the problem with those was that they were taking too long, and/or using too much memory to be used as regular test suite.
Frank
Users mailing list Users@einsteintoolkit.org http://lists.einsteintoolkit.org/mailman/listinfo/users
users@lists.einsteintoolkit.org
-
Bernard Kelly -
Erik Schnetter -
Frank Loeffler