#78 -timeout N precludes timeout condition

closed-invalid
None
5
2010-08-31
2010-07-22
David T Kerns
No

if I specify a -timeout N on an expect (to override the global timeout) then the timeout condition is not executed. It does properly timeout.
running version 5.43.0 on Linux

Discussion

  • David T Kerns
    David T Kerns
    2010-07-22

    demonstrates bug on *nix OS

     
    Attachments
    tt
    • assigned_to: nobody --> andreas_kupries
     
  • Hm. I actually see the bug with an 5.43 build on linux as well ...

    It is apparently not the -timeout 1 per se, but the fact that its presence prevents the interpretation of the argument after it as pattern/action list. It seems to be interpreted as a plain pattern with an empty action.

    Continuing the investigation ...

     
  • I have now gone back to 5._41_.0 (Jun 7, 2004) and it exhibits the proclaimed bug. I am on linux. I would like to know where the 5.43 comes from which doesn't have the bug, as I wonder if the vendor applied local patches and fixes.

    It is definitely the command line processing which recognizes a single pattern/action list argument as such if and only if it is (a) the one and only argument of the 'expect' command, or (b) it is preceded by a '-brace' flag, and flag + list are the sole two arguments of the 'expect' command. The moment additional flags are provided the pattern/action list is treated as pattern argument. And as the last argument it has no action.

    The sources as far back as Jun 7, 2004 bear this out.

    This long-evity of the behavior, plus the documentation I am not fully convinced that what you see is a bug. The way the documentation (*) is written I would actually expect (sic) the '-timeout 1' arguments to be either inside of the pattern/action list (**), not outside, or to have everything as separate arguments.

    And indeed, when doing either the 'expect' command uses the local timeout. I am getting convinced that the bug is in pseudo-bad syntax of the invokation, not the expect package. Pseudo-bad because expect has a proper interpretation for the argument, just not the one you expected.

    (*) http://docs.activestate.com/activetcl/8.5/expect/expect.1.html
    (**) " expect [[-opts] pat1 body1] ... [-opts] patn [bodyn] "

    Another issue, the text '# expect this line' is NOT a comment, but the patterns '#' and 'this' with actions 'expect' and 'line'. Well, when the argument is recognized as pattern/action list.

     
  • My revision of tt with report of version and location of loaded package, and the confusing date info replaced by fixed strings.

     
  • Revised script demonstrating one form of correct syntax (everything as separate arguments)

     
  • Revised script demonstrating the second form of correct syntax (the options in the pattern list)

     
    • status: open --> closed-invalid