Re: [lmi] assert in stratified_entity::assert_validity()

[Greg Chicares]

On 2010-07-25 22:48Z, Vadim Zeitlin wrote:
> 
>  If I start a LMI and do Alt-F,N,I to create a new illustration I get an
> assert in stratified_entity::assert_validity() because limits_.back() is
> 1.0 and not infinity<double>(). The precise stack at the moment of assert
> is
> 
>       stratified_entity::assert_validity()  Line 136
>       stratified_entity::read(const xml::node & e={...})  Line 164

At that point, it should have finished reading an xml element,
so the object should pass all validity tests.

> Would you know of an explanation as to why this happens by chance? I did
> regenerate the project files, including sample.strata from which the
> offending value is being read by stratified_charges ctor, using
> product_files. Should I do something else to fix this?

That's the right way. Every '</limits>' should be preceded by
'<item>inf</item>':

$grep -B1 '</limits>' sample.strata |sed '/item/!d' |uniq; echo END
      <item>inf</item>
END

> Or is it some
> artefact of MSVC build (I didn't test under Linux yet because I still
> didn't finish my floating point changes there)?

Oh...msvc...and you observe a value equal to one...sounds like the
problem I found with the OS's 'msvcrt.dll'. I figured they would
have fixed that in their commercial compiler, which provides its
own updated runtime.

'numeric_io_test.cpp':
    // This test fails for como with msvcrt, because the latter
    // defectively prints infinity as "1.#INF". Distressingly,
    // that converts to '1.0'.
    BOOST_TEST_EQUAL(inf_dbl, numeric_io_cast<double>(inf_str));

ISO/IEC 9899:1999 [7.19.6.1/8]:
| A double argument representing an infinity is converted in one of the
| styles [-]inf or [-]infinity — which style is implementation-defined.