Open StephanWatkins opened 2 weeks ago
I decided to update the csh and tcsh on a PC at work I had gamess running fine on before with just the normal rungms xxxfile 1 1 >& log.log & command from a normal bash shell, compiled with simple sockets on 6 core i5 machine. After updating (which I think the last was a year 1/2 back when it was installed) I now get the same error message from bash, or from csh as above?
on installing csh for bash:
Preparing to unpack .../csh_20110502-7_amd64.deb ... Unpacking csh (20110502-7) ... Setting up csh (20110502-7) ... update-alternatives: using /bin/bsd-csh to provide /bin/csh (csh) in auto mode Processing triggers for man-db (2.10.2-1) ...
I noticed it said using bsd-csh although I have no clue what it was using before...
I found a way to get it to run, which I will not try and figure out, but will post in case someone runs into this. Something changed in the newer csh for bash in the Ubuntu repositories (also used for linux mint, including generic kernels).
tcsh rungms xxxfile 00 1 1 0 0 >& log.log&
:or variants with tcsh as the command line initiation from a bash shell, why csh, or just running it alone in the bash shell no longer work I do not know. Also, if you drop to a tcsh or csh shell, it no longer works, and gives the above errors.
Check the first line of the 'rungms' script. It looks like you messed it up. It must be
#!/bin/csh
without any space at front.
I changed it temporarily to the share bin csh, but it normally most times I tried it was #!/bin/csh... I have it running now, but have to type tsch, in my laptop I have to type csh? I think it was something changed in the csh for bash they started using an alternate csh rather than just csh. The laptop has ubuntu 20...
Stephan
On Tue, Apr 29, 2025 at 9:19 AM Ivan Rostov @.***> wrote:
ivr900 left a comment (gms-bbg/gamess-issues#82) https://github.com/gms-bbg/gamess-issues/issues/82#issuecomment-2838475108
Check the first line of the 'rungms' script. It looks like you messed it up. It must be
!/bin/csh
without any space at front.
— Reply to this email directly, view it on GitHub https://github.com/gms-bbg/gamess-issues/issues/82#issuecomment-2838475108, or unsubscribe https://github.com/notifications/unsubscribe-auth/BCLVIBYOYDAKVYJK52KR7MD235UXTAVCNFSM6AAAAAB37KV2Q6VHI2DSMVQWIX3LMV43OSLTON2WKQ3PNVWWK3TUHMZDQMZYGQ3TKMJQHA . You are receiving this because you authored the thread.Message ID: @.***>
There is nothing special in Ubuntu that would prevent running Tcsh scripts from bash. As with any linux/unix system, three conditions are required for that:
Only if one of the tree conditions above is not satisfied can explain the symptom you described, i.e., scripts works as argument of the tcsh command but not running on its own.
Like I said it runs if I type tcsh before the line, and in the past I have a laptop that requires csh typed before, in all, the scripts, permissions and interpreter switch are fine at the top, I can not for the life of me figure out why the differences are there for execution?
On Thu, May 1, 2025 at 10:46 PM Ivan Rostov @.***> wrote:
ivr900 left a comment (gms-bbg/gamess-issues#82) https://github.com/gms-bbg/gamess-issues/issues/82#issuecomment-2846123579
There is nothing special in Ubuntu that would prevent running Tcsh scripts from bash. As with any linux/unix system, three conditions are required for that:
- path to the script provided (note if you run a script from the same directory the command must start as ./rungms unless . in the PATH, which is not recommended in general)
- proper execution permissions of the script, e.g., chmod u+x rungms
- a proper first line of the script. It ust be "#!/bin/csh" or "#!/bin/env csh" without any extra characters in front or after. Some special characters, which are not be visible in some editors, may get there by mistake or during file transfer, therefore if not sure, try to delete the line and write it again.
Only if one of the tree conditions above is not satisfied can explain the symptom you described, i.e., scripts works as argument of the tcsh command but not running on its own.
— Reply to this email directly, view it on GitHub https://github.com/gms-bbg/gamess-issues/issues/82#issuecomment-2846123579, or unsubscribe https://github.com/notifications/unsubscribe-auth/BCLVIBZLIMDT2SUSYH7MGEL24LFABAVCNFSM6AAAAAB37KV2Q6VHI2DSMVQWIX3LMV43OSLTON2WKQ3PNVWWK3TUHMZDQNBWGEZDGNJXHE . You are receiving this because you authored the thread.Message ID: @.***>
High,
Wondered if this was a bug with Gamess, or something on my OS. I can't find a problem on my OS with the bash or csh shells.
I have a bizarre error, after switching and installing on a new pc. I had gamess running on a 6 core pc, and 6 core laptop, but recently switched to a new 24 core pc. I tried installing 3 different binaries in different locations with mkl, openblas and the simple sockets and internal lappack. The error I am very familiar with, it's a simple bash/csh error that says your script has a character or variable not allowed:
"Illegal variable name."
However I have tried using a bash, csh, tcsh and zsh shell, and get the same error. If I run the ./runall check from the software I get"
1 : exam49 2 : 02 3 : 1 Illegal variable name. " The same if I use rungms to check one of the examxx inp files. I have tried a dozen variations of launching the rungms as well, direct path, launching in the gamess dir, as a bash or csh .sh, .csh file with parameters (file 02 2 1 00, or 1 1, 2 1 , 1 1 0 etc...)
If I run one of the gamess binaries ./gamess.02.x as an example I get the standard error: " ./gamess.02.x
Error: Expecting an even number of MPI processes (cp:ds::1:1). DDI Process 0: error code 911
MPI_ABORT was invoked on rank 0 in communicator MPI_COMM_WORLD with errorcode 911.
NOTE: invoking MPI_ABORT causes Open MPI to kill all MPI processes. You may or may not see output from other processes, depending on exactly when Open MPI kills them.
"
or ./gamess.02.x 2 1
" Program received signal SIGSEGV: Segmentation fault - invalid memory reference. ....
9 0x591f0f6c0fb0 in beging_
10 0x591f0f6bdb6c in gamess
11 0x591f0f6b00fe in main
Segmentation fault (core dumped)
"
or for the sockets version just missing input and....error
Indicating the binaries are working fine, at least as far as accepting inputs..
I for 2 days have been pulling out my hair trying to figure out where the error comes from, and checked my bash or csh shells, which can run scripts fine, so I assume something with the gamess, either in compile or the rungms itself. It has all permissions set to owner and execute. I installed the latest src, and tried the r2 july 15th. I am on Linux Mint 21.3, intel i9-14900KF 24 core, 5070 Ti, although I did not try compiling the CUDA version yet, but have not had any problems with anything else. Anyone have an idea? I mean it seems like it is not taking the arguments into the rungms from command lines, the rungms run alone with no input just gives the "illegal variable name" error, and I have checked a dozen times the settings inside matched against what I had on the other PCs, and only changed the first 4 lines, to the appropriate directories, and the 2 areas for openmpi and openmpi library location as per instillation instructions. Compiling with sockets gives the same error, as mentioned, mint is the 6.8.0-58, latest kernel. Of note, on my laptop install from 1 1/2 years back i have to csh before rungms, on the PCs this is not necessary. I also tried xterm, term terminals as variants.
Stephan Watkins