Re: [boost] Review Request: regular expressions

From: John Maddock (John_Maddock_at_[hidden])
Date: 2000-07-05 06:44:34

Jens,

Thanks for prompt response, pity about the error's!

>I would be pleased if the #include files, particularly those in
boost/detail, could have more verbose names.  The 8.3 naming
convention should be abolished.  Oh, and a common prefix (subdir?)
would be cool, too.
<

Yep, will take a while though.

>lib/makefile needed hand-fixing because it contained some CR/LF
>line-endings, which upset my GNU make 3.77

OK, I'll update my zip build script, note that the zip -a option will fix
this though...

>src/c_regex_traits.cpp fails because re_strlen and re_strcpy are
>unknown. They are used there in an anonymous top-level namespace,
>but are defined in boost::re_detail. Appropriate explicit
>prefixes solve the problem.
>
>src/c_regex_traits.cpp accesses regex_message_catalogue in a
>global function, although it is a member of c_traits_base.
>

I can see how I introduced those errors when rewriting that section - but
how on earth did *all* my test compilers miss this?  Deeply worrying.

>Results for gcc 2.95.2 on Linux/x86, Dietmar Kuehl's cxxrt:
>
>C header file wrapping doesn't quite work, so I get some identifier
>conflicts.  Given up.

If you had any idea how much hassle I've had from broken C headers....
still it's arguably not my problem... if there are specific fixes though
let me know, but the C header wrappers are probably the right thing to fix
(in an ideal world).

>Results for gcc 2.95.2 on Linux/x86, STLport 4.0b8:
>
>There seems to be a problem with wide character support in
>my glibc 2.1.2.  In particular, I haven't found fgetwc
>and friends, yet.  Given up for now.
>

I'm not sure where those come from, I certainly don't ever use fgetwc in
regex++, looks like STLPort needs patching. BTW it's possible to turn off
wide character support using:

configure --disable-wstring

and "configure --help" gives the full list of options.

As an aside I don't think that I have ever come accross a platform with
conformant wide character headers - not even conformant with the C standard
never mind C++, sometimes I dispare at these ever being fixed.

Anyway, I'm uploading a patched version now, hopefully it should be legal
C++ this time (!), how far it will take you with wide character support is
another matter...

Thanks for the feedback,

- John.