Unable to run Gabriel benchmarks on Linux

[pamoroso]

To run the Gabriel benchmarks I cloned Kabir's branch of the gabriel repo and went through the steps under Quick Start in the README. When I executed the runner script in step 4 I got a X11 display error:

paolo@lispmachine:~/medley/gabriel$ ./do-bench.sh bench-1
Running bench-1
  BENCHDIR   = /home/paolo/medley/gabriel/Benchmarks/BENCH-1
  RESULTSDIR = /home/paolo/medley/gabriel/Results/BENCH-1
  CMFILE     = /tmp/benchmarks-52418/hcfiles.cm
"/home/paolo/bin/lde" "/home/paolo/medley/medley/loadups/apps.sysout" -id "bench_1" -title "Medley::bench_1" -g 1024x768 -sc 1024x768 -noscroll
MEDLEYDIR: "/home/paolo/medley/medley"
LOGINDIR: "/tmp/benchmarks-52418"
GREET FILE: "/home/paolo/medley/medley/greetfiles/NOGREET"
REM.CM FILE: "/tmp/benchmarks-52418/hcfiles.cm"
VMEM FILE: "/tmp/benchmarks-52418/vmem/lisp_bench_1.virtualmem"
Unable to open X11 display from DISPLAY
paolo@lispmachine:~/medley/gabriel$ echo $DISPLAY
:0

Is there an issue with my setup? I'm on Linux Mint 22.1 Cinnamon under X11 and run Medley built for X11.

Also, section Results says:

Important: do-bench.sh currently loads NEW-BENCH-*.LCOM for all suites. To get .txt output in actual runs, recompile each NEW-BENCH-* source in Medley so the updated output names are included in the .LCOM files.

It's not clear to me when the NEW-BENCH-* sources are supposed to be recompiled, prior or after running ./do-bench.sh bench-*.

[hjellinek]

I'm rewriting and reorganizing Kabir's product. I can address both of your points.

1. Regarding your X11 problem, I can say that when I tried Kabir's branch before I started working on it, it ran fine under SDL. Make of that what you will.

Important: do-bench.sh currently loads NEW-BENCH-*.LCOM for all suites. To get .txt output in actual runs, recompile each NEW-BENCH-* source in Medley so the updated output names are included in the .LCOM files.

It's not clear to me when the NEW-BENCH-* sources are supposed to be recompiled, prior or after running ./do-bench.sh bench-*.

I checked with Kabir. He compiled the NEW-BENCH-n source files, checked in the compiled files, changed the code to make the results files' extension .txt, and checked in the updated sources without compiling them. Thus the instruction to compile them, meaning you should do that before running.

However, if I understood his email correctly, NEW-BENCH-n are the only files he ever compiled, which supports @masinter's theory that the bugs he encountered and reported were due to executing code that was compiled for an earlier virtual machine.

I recommend you not devote further effort to Kabir's branch. I hope to have a replacement before too long.