Work at SourceForge, help us to make it a better place! We have an immediate need for a Support Technician in our San Francisco or Denver office.


#60 run form fails in restored image

run-time (53)
david rush

I have a test program which runs an external
program using a (run ...) form with redirection.
When I run from a saved image I get the following
run-time errors [and an image crash]:

drush@dubcvs: test foo bar
heap size 4000000 is too small, using 4361408
undefined imported name sch_tcflow
undefined imported name sch_isatty
undefined imported name sch_tcdrain
undefined imported name format_date
undefined imported name scm_ctermid
undefined imported name sch_tcflush
undefined imported name sch_ttyname
undefined imported name sleep_until
undefined imported name sch_tcgetpgrp
undefined imported name sch_tcsetpgrp
undefined imported name char_ready_fdes
undefined imported name sch_tcsendbreak
undefined imported name scheme_tcgetattr
undefined imported name scheme_tcsetattr
undefined imported name time2date
undefined imported name open_ctty
Long distance runner: foo bar
Error while running root thread, thread killed: #{Thread 3}
Exception: 33
#{Procedure 11533 (unnamed in unnamed in unnamed in %wait-any in scsh-level-0)}
(#f 14499 0)

Error: returning wrong number of values
(#{Procedure 11554 (unnamed in really-wait in scsh-level-0)} 10 -1 0)

Obviously the undefined imported names are not
interesting, but the wrong-number of arguments
and returning wrong number of values *are*

I am using scsh-0.6 built from the head of the CVS
tree on November 5.


1 2 > >> (Page 1 of 2)
  • david rush
    david rush

    Logged In: YES

    I have a tarball of the files which I have been attempting
    attach to this bug w/out much success. I've just changed
    browsers in the hope that's the problem, but I can email
    it whenever necessary...

    If it has uploaded you need to untar and run tb.script. then
    run the program test as above...

  • david rush
    david rush

    tarball of files isolating the bug

  • Logged In: YES

    Sorry, but for me the script is running fine:

    aubisque[63] ./test foo bar
    Long distance runner: foo bar
    aubisque[64] less bar
    this is a fake...
    sablotron args:

    I think something with your build is wrong. The undefined
    imported names also point into this direction. Another
    problem is that the error messages look rather strange, they
    seem to be out of sync. Is it possible that they originate
    from the forked process?

    • assigned_to: nobody --> mainzelm
  • david rush
    david rush

    Logged In: YES

    Then how can I further diagnose/fix the problem? There are
    messages originating from the program run in the subprocess;
    whether or not Solaris underneath it all is complaining, I
    no good way of diagnosing.

    Could I be looking at some problems in my initial image?

  • Logged In: YES

    Did you redo the whole build process, including a cvs
    update? Apart form that I'm a little bit helpless here. I
    checked the source but could not find a place that would
    match your error messages. I also tried your script on
    Solaris 5.7, gcc 2.96 but did not have any problem.

    Can you reproduce your error in the interactive mode and
    gather some debugging information using ,debug?

  • david rush
    david rush

    Logged In: YES

    I followed the instructions in CVS_README:

    set BUILD_RUNNABLE to local Scheme48-0.53

    The error does *not* occur in interactive mode. That is why
    submitted a bug report. I will try a rebuild and capture the
    output for closer examination...

  • Logged In: YES

    I'm now almost sure that your scsh/syscalls1.c is not
    up-to-date. In revision 1.22 wait_pid returned a list of
    three elements, now in 1.23 there are only two elements.
    This list is turned into multiple values in the Scheme code.

  • Logged In: YES

    David, can you please check whether you use an outdated version of the VM to launch the image?

  • david rush
    david rush

    Logged In: YES

    Yes, I have just updated my CVS build (Nov 14) and the
    bug appears to be gone.

    Close it , Dano!

1 2 > >> (Page 1 of 2)