Compiling instructions for linux

[SSchott]

Hi,
I was checking if there are any updated instructions for compiling clasp/cando in an "agnostic" linux distro, also without being root? The deb packages have been compiled with glibc 2.34, which is not available in most stable releases. I managed, by digging into the homebrew recipes and by using conda with the clasp repo to get it compiled, but doesn't work properly; calling cando results in a lost terminal, irrespective of what arguments I pass:

Starting clasp-boehmprecise-2.3.0-48-gc050e511e-ga4c1ad67 from extension image
Loading resource file /home/schott/.clasprc
To load "cando-user":
  Load 1 ASDF system:
    cando-user
; Loading "cando-user"

To load "seqan":
  Load 1 ASDF system:
    seqan
; Loading "seqan"

Creating logical host AMBER with a root path of /apps/prod/amber/23.1/
Paths: (/home/schott/ /apps/prod/amber/23.1/dat/antechamber/
        /apps/prod/amber/23.1/dat/leap/cmd/oldff/
        /apps/prod/amber/23.1/dat/leap/cmd/
        /apps/prod/amber/23.1/dat/leap/parm/
        /apps/prod/amber/23.1/dat/leap/lib/
        /apps/prod/amber/23.1/dat/leap/prep/)
Creating LPARALLEL kernel with 16 threads.
Top level in: #<PROCESS TOP-LEVEL @0x7fa800522f09 (Running)>.
CANDO-USER> help

Condition of type: UNBOUND-VARIABLE
The variable CANDO-USER::HELP is unbound.
Available restarts:
(use :r1 to invoke restart 1, etc.)

1. (RESTART-TOPLEVEL) Go back to Top-Level REPL.

ERROR
COMMON-LISP-USER>> 

[SSchott]

Also while trying to execute scleap (even in pop-os), I get the following:

/usr/share/clasp/src/llvmo/code.cc:261:fixupInternalsForSnapshotSaveLoad for Library_O - resetting pointers for DWARFContext - make sure this works
/usr/share/clasp/src/core/lisp.cc:2081 -- *nested-error-depth* --> 11  datum: SIMPLE-TYPE-ERROR
               initializers: (:DATUM COMMON-LISP::SIMPLE-TYPE-ERROR :EXPECTED-TYPE COMMON-LISP::CONDITION :FORMAT-CONTROL "Not a condition type: ~S" :FORMAT-ARGUMENTS (COMMON-LISP::SIMPLE-TYPE-ERROR ))

[yitzchak]

The building instructions are here. You'll need to add the cando extension as detailed there also. https://github.com/clasp-developers/clasp/wiki/Building-and-Installing-from-Source

[SSchott]

Thanks for the prompt reply.
I tried pretty much that in an OpenSuse installation, with the results in the first message. The latter message applies both for an attempt in OpenSuse, and in Pop-OS (using the deb package repo available instead of compiling from source for Pop-OS).
Any ideas, at least for the Pop-OS case?

[yitzchak]

I'm a little confused by your first post. There is no "HELP" command in the Lisp REPL. Are you thinking of the LEAP command?

Also I don't know what you mean by a lost terminal. Can you clarify?

[SSchott]

Yes, I was indeed thinking on the usage of cando, never used it before, so I am a bit confused myself. Using :exit works just fine, I guess I have to find documentation on usage.

Still, the second error appears either by callin scleap or by trying to generate a console with jupyter-lab, as suggested in the demos (again, with Pop-OS, as I want to see if I manage to get things working with the least resistance path first).

[yitzchak]

If you want a LEAP terminal try the cleap binary. scleap is a snapshot version. Basically, some stuff precompiled for speed.

Can you post the full log from the scleap failure?

[SSchott]

Ok, thanks for the pointers, it is a bit blurry what is what. cleap itself runs, even in my dodgy OpenSuse compilation. It broke while loading the lipid21 FF though, because of the lower #### comment section. I made a local copy within the folder without it and I am checking if it goes through one of my parametrization runs.

Here is the scleap error:

usr/share/clasp/src/llvmo/code.cc:261:fixupInternalsForSnapshotSaveLoad for Library_O - resetting pointers for DWARFContext - make sure this works
/usr/share/clasp/src/core/lisp.cc:2081 -- *nested-error-depth* --> 11  datum: SIMPLE-TYPE-ERROR
               initializers: (:DATUM COMMON-LISP::SIMPLE-TYPE-ERROR :EXPECTED-TYPE COMMON-LISP::CONDITION :FORMAT-CONTROL "Not a condition type: ~S" :FORMAT-ARGUMENTS (COMMON-LISP::SIMPLE-TYPE-ERROR ))
Dumping backtrace
/usr/share/clasp/src/core/debugger.cc:614:dbg_safe_backtrace This is where we would have clasp dump a backtrace that doesn't use the printer or allocate memory
          We don't currently have that functionality - but if we did it would use backtrace(), backtrace_symbols()
          and scrape the DWARF accessible through _lisp->_Roots._AllObjectFiles to build a backtrace with JIT function names
          we can use the ObjectFile_O and Code_O objects to figure out what return addresses are JITted code and what DWARF
          belongs to each return address.
          It might be better to use the gdb JIT API (https://sourceware.org/gdb/current/onlinedocs/gdb/JIT-Interface.html)
          What I'm missing to make the gdb JIT API work is how do we know where the code lives for the object file/symfile?
          It would be useful to examine __jit_debug_descriptor and look at the symfile_addr/symfile_size entries to figure out
          where they point relative to what we have in _lisp->_Roots._AllObjectFiles.
          As of May 2, 2021 only ELF files work with the gdb JIT API - so this wouldn't work on macOS
Aborted (core dumped)