Re: Assertion in symbol table due to Revision 8881

[Jaroslav Hajek]

On Mon, Jun 29, 2009 at 8:10 AM, Daniel J Sebald<address@hidden> wrote:
> John,
>
> I've come across an assertion that I traced back to the first (second?) mod
> following version 3.1.53:
>
> --- a/src/ChangeLog     Thu Feb 26 12:25:22 2009 -0500
>      2 +++ b/src/ChangeLog     Thu Feb 26 13:58:47 2009 -0500
>      3 @@ -1,5 +1,25 @@
>      4  2009-02-26  John W. Eaton  <address@hidden>
>      5        6 +       * symtab.h
> (symbol_table::symbol_record::symbol_record_rep::forced):
>      7 +       New static constant.
>      8 +
> (symbol_table::symbol_record::symbol_record_rep::force_variable,
> ...
>
> labelled as [Thu, 26 Feb 2009 13:58:47 -0500] rev 8881.  (Wish Hg-viewer
> simply listed the date rather than a vague "4 months ago"... BTW, without
> the ChangeLog file I likely would have given up trying to locate where the
> bug appeared.)
>
> I'm afraid I can't be much help with a short example that causes the
> assertion.  The code segment is embedded in a bigger convoluted program and
> the script file is a recursive one.  Small examples I've tried don't cause
> the problem on its own.  The best way for me to describe the occurence is
> "An assignment made inside of an eval(), inside of a nested/recursive script
> routine."
>
> The octave-core dump ends up being only 11 bytes long.
>
> The actual note is
>
> octave: symtab.h:1066 : static void symbol_table::force_variable(cost
> std::string&, int, size_t): Assertion
> `xcurrent_context == 0' failed.
> panic: Aborted -- stopping myself...
> attempting to save variables to 'octave-core'...
> save to 'octave-core' complete
> Aborted
>
> Version 3.1.53 works fine.
>
> Dan
>
>

Hmm, I don't see why the assertion is valid. Surely force_variable
*can* be executed in nonzero context (i.e. recursive call), using
eval, as your example shows.

Another example is

function testf2(n)
if (n > 0)
  testf2 (n-1);
else
  eval(["a = ", num2str(n)]);
endif


octave:1> testf2(3)

Since it doesn't seem to break anything, I've removed the assertion:
http://hg.savannah.gnu.org/hgweb/octave/

thanks

-- 
RNDr. Jaroslav Hajek
computing expert & GNU Octave developer
Aeronautical Research and Test Institute (VZLU)
Prague, Czech Republic
url: www.highegg.matfyz.cz