Re: [lmi] [PATCH] wx_test_validate_output.cpp

[Greg Chicares]

On 2015-03-10 16:58, Vadim Zeitlin wrote:
> On Mon, 09 Mar 2015 21:25:37 +0000 Greg Chicares <address@hidden> wrote:
> 
> GC> On 2015-03-03 22:51, Vadim Zeitlin wrote:
> GC> > On Mon, 02 Mar 2015 00:16:05 +0000 Greg Chicares <address@hidden> wrote:
> GC> > 
> GC> > GC> I've copied my results below. I expect
> GC> > GC>   default_update: ERROR (Assertion failure: A message box "Unable 
> to read file 'C:/etc/opt/lmi/default.ill'. [file 
> /lmi/src/lmi/illustration_document.cpp, line 110] " was shown unexpectedly, 
> expected defaults dialog. [file /lmi/src/lmi/wx_test_default_update.cpp, line 
> 96, in run()].)
> GC> > GC> until that test is made conditional on '--distribution',
> GC> > 
> GC> > [As an aside, I'll post the long due patch doing this soon]
> GC> 
> GC> Okay, that one no longer occurs.
> GC> 
> GC> However, I see this new error:
> GC> 
> GC> input_sequences: ERROR (Input validation problems for '': Interval [ 20, 
> 310 ) is improper: it ends after the last possible duration. Current token 
> ';' at position 42.  [file /lmi/src/lmi/input_realization.cpp, line 260] )
> GC> 
> GC> The default input and retirement ages are 45 and 65 respectively,
> GC> as I have no default input file. The '310' doesn't make sense.
> GC> I never saw anything like that with the old version.
> 
>  I can't reproduce this here, even after deleting the default file
> (/cygdrive/c/fop-0.20.5/coli_boli_default.ill if I understand correctly).
> And it's very worrisome that we don't get the same results, there must be
> some difference in our configuration that I completely fail to see.

My theory is that both problems reported in this message are related to
the wx action-simulation code, perhaps in combination with msw-xp. You
recently said it was substituting hyphen '-' for underscore '_'....

I added debug-print statements and found that this sequence fails:

"SpecifiedAmount" ,"sevenpay 7; 250000 retirement; 100000 #10; 75000 @95; 50000"

With my default input and retirement ages, 45 and 65 respectively, the
retirement duration is twenty. The sequence means
  sevenpay for seven years; then
  250000 until duration twenty; then
  100000 from duration twenty...until..."#10" means ten years later...
    ...which would be [20, "#10")
    ...which should be interpreted as [20, 30)...

but I get

  Interval [ 20, 310 ) is improper

Now type "#10" without holding down shift for "#": you get "310",
at least on a US keyboard.

This is the first sequence, and the test abends at the first failure,
so there may be further problems in the later input-sequence tests.

> GC> > GC> but is this error:
> GC> > GC>   validate_output_census: ERROR (Assertion 'output_values.exists()' 
> failed (file "C:/opt/lmi/bin/values.tsv" after print case to PDF). [file 
> /lmi/src/lmi/wx_test_validate_output.cpp, line 360] )
> GC> > GC> expected?
> GC> 
> GC> I still have that error.
> 
>  And I still don't :-( Just to verify: do you only get these errors when
> running all the tests (which could indicate some spooky bug depending on
> the exact timing of something) or even when you just run the individual
> tests, i.e. "./wx_test.exe -t input_sequences" or "./wx_test.exe -t
> validate_output_census"? FWIW I typically run just the individual tests,
> although I did run the full test suite before replying to this message.
> I also use --data_path and --a......g arguments.

Same outcome whether I run the tests individually or all together.

> GC> Within five minutes of rerunning the test suite I did this:
> GC>   find /cygdrive/c -mmin -5 | grep tsv
> GC> It found
> GC>   /cygdrive/c/opt/lmi/bin/testfile.mec.tsv
> GC> and several 'MSEC*' files in the same directory, but no "values.tsv".
> ...
> GC> /// Census | Edit case defaults
> GC> ///   Comments: replace contents with "idiosyncrasy_spreadsheet"
> GC> ///   OK
> GC> ///   Yes (apply to all)
> GC> /// Census | Print case to PDF
> GC> /// Verify that this file was created:
> GC> ///   values.tsv
> GC> 
> GC> Now /opt/lmi/bin contains "values.tsv" along with three pairs of
> GC> new "ABC*.fo.xml" and "ABC*.pdf" files.
> 
>  The combination of these facts leads me to believe that the automatic test
> didn't enter the special comment value correctly into the dialog. If this
> is true, the attached patch should at least detect this and provide a more
> clear error message -- could you please apply it locally and try running
> with it?

Hyphen substituted for underscore:
  idiosyncrasy_spreadsheet <-- what it should enter in "Comments"
  idiosyncrasy-spreadsheet <-- what it does enter instead

[I would copy and paste the actual message, but I've disabled the
'spice' tools that would let me do that, because I strongly suspect
they were responsible for frequent VM lockups.]