Stephen, and Ulrich,
The cintltst or intltest tests will fail with VC7 in release mode, but
not in debug, as Ulrich will find out. This is because of a compiler bug
we ran into with regards to VC7 only. Over the weekend this jogged my
memory a bit and I thought of something Ulrich could try. There is a
patch for ICU that was issued by IBM for a VC7 compiler bug that caused
some problems with the release builds.
In particular, make sure that if you are using VC7 that you apply the
patch associated to version 1.26. Then recheck your results. This will
at least get the tests working properly.
Ulrich, I realize this is probably a dumb question on my part, but are
you also using the VC7 project files that are part of the distribution,
or ones converted from VC6 by yourself? The reason I ask is because in
the process of converting the projects to VC7 I did quite a bit of
cleanup, and made things much more consistent. I'd hope that you'd be
able to take advantage of that.
> -----Original Message-----
> From: icu4c-support-admin@...
> [mailto:icu4c-support-admin@...] On
> Behalf Of srl@...
> Sent: Friday, September 27, 2002 4:13 PM
> To: Ulrich Straub
> Cc: icu4c-support@...
> Subject: Re: Problem with number parser and VC7 .NET
> Do cintltst or intltest tests pass in debug mode (non optimized, I
> On Friday, September 27, 2002, at 09:39 AM, Ulrich Straub wrote:
> > I got a problem with ICU 2.0 and ICU 2.2 (other versions
> not tested)
> > compiled with VC7 when trying to parse a number string (for example
> > convert the string "644" to the number 644).
> > The problem can be recreated by compiling the ICU sample
> "numfmt" with
> > VC7
> > using
> > the following makefile: ...
> > CFLAGS=/c /MT /Zp1 /W3 /GX /nologo
> > #CFLAGS=/c /MT /Zp1 /W3 /GX /nologo /O2
> > When compiling with the option /O2 the sample works.
> > Question:
> > --------------
> > Does this mean there is a bug in ICU, since the code should
> work also
> > without optimization ?
> > Based on this result I have the feeling that the ICU code is quite
> > fragile.
> > Is this true or is there a simple explanation why ICU with VC7
> > requires the
> > option /O2 ?