Fix dummy argument aliasing issue #17
Conversation
|
@whlipscomb - can you please review this and let me know if it looks okay? Then I'll need to rebase this to a common ancestor of the master and release branches and merge it into both. |
NAG version 6.2 catches a problem with passing the same variable to two different arguments of a subroutine, one of which is intent(out). The fix is to use a separate copy of the local variable for one of the arguments.
|
I have rebased onto the release branch and force pushed. The diffs are the same as before. |
As with the previous commit: NAG version 6.2 picked up some of these issues.
|
@gunterl okay, you can go ahead and test these changes. I'm going to restart my own testing, too. |
|
sounds good.
…On Fri, Jul 6, 2018 at 4:24 PM, Bill Sacks ***@***.***> wrote:
@gunterl <https://github.com/gunterl> okay, you can go ahead and test
these changes. I'm going to restart my own testing, too.
—
You are receiving this because you were mentioned.
Reply to this email directly, view it on GitHub
<#17 (comment)>, or mute
the thread
<https://github.com/notifications/unsubscribe-auth/AH0K-AQDmAca1nPUYywBWTj0BAOYe3MOks5uD-OvgaJpZM4VFp3X>
.
|
|
I have completed my cism-in-cesm testing with these changes, and everything looks good: The NAG tests pass, and the cheyenne-intel/gnu tests pass and are bit-for-bit. @gunterl and @whlipscomb once you sign off on these changes I'll bring the changes to master and the release branch. I want to do this in a very specific way to maintain the history I used for my testing in CESM, so please don't merge this yourself. |
|
Sorry: While I'd like to wait for you two to sign off on this, I'm actually going to go ahead and merge this this weekend: The git steps I need to do are complicated enough that there's a good chance I'm going to mess something up if I wait to do this until I'm back from vacation. (This is complicated by the facts that (1) these fixes are needed ASAP for CESM testing, both on master and the release branch; (2) I'm trying to keep master, release and release-nohist all in sync.) So you'll probably see this as merged soon (maybe tomorrow). I'll still welcome a review of these changes, though. If you see any problems, we can fix them in a separate commit. |
|
@billsacks, I looked at the changes in parallel_mpi.F90 and glissade_transport.F90, and they seem fine. I also checked out escomp/master (after you merged these changes) and ran some standard standalone tests that (as I expected) had no problems. So I'm ready to sign off. Thanks for taking care of this fix. |
billsacks
changed the title
|
Hi Bill,
I was about to run the stand alone simulation this afternoon but it looks
like everything seems to be in working order. Thanks for dealing with this.
Gunter
…On Sun, Jul 8, 2018 at 11:25 AM, Bill Sacks ***@***.***> wrote:
Merged #17 <#17>.
—
You are receiving this because you were mentioned.
Reply to this email directly, view it on GitHub
<#17 (comment)>, or mute the
thread
<https://github.com/notifications/unsubscribe-auth/AH0K-OIrBSOd4SZWjyc7sEZ9WhYkQVZGks5uEkCggaJpZM4VFp3X>
.
|
Allow more characters in message string This is a very simple PR to fix a problem seen by the Delft group. The code was failing in glimmer_nc_readparams when the restart filename was more than ~70 characters (e.g., because of a long absolute path). The fix is to allow more characters in the 'message' string.
NAG version 6.2 catches a problem with passing the same variable to two
different arguments of a subroutine, one of which is intent(out). The
fix is to use a separate copy of the local variable for one of the
arguments.
Fixes #6