Regardless of the __i keyword problem, the internal error is a compiler bug. Could you zip up a test case and email it to me please? walter@digitalmars.com



"Damian Dixon" <damian.dixon@tenetdefence.com> wrote in message news:1105_984746733@dilbert...
> I've now run into a full stop..... I've changed the '__i's to 'i__i' and
now:
>
> sc
console.cpp -mn -C -WA -S -3 -a8 -c -gf -D_CONSOLE=1 -Ab -Id:\sc\stltemp\STL \port\stlport -oconsole.obj
> sc
regression\string1.cpp -mn -C -WA -S -3 -a8 -c -gf -D_CONSOLE=1 -Ab -Id:\sc\ stltemp\STL\port\stlport -oregression\string1.obj
> Fatal Error: D:\SC\stltemp\STL\port\test\regression\string1.cpp(32):
internal error cg87 1144
> Lines Processed: 16360  Errors: 1  Warnings: 0
> Build failed
>
> I will have to look at a small stand alone test :<
>
> Any suggestions would be welcome.
>
> On Fri, 16 Mar 2001 12:40:49 GMT, Damian Dixon
<damian.dixon@tenetdefence.com> wrote:
> > On Fri, 16 Mar 2001 01:12:44 -0800, "Walter" <walter@digitalmars.com>
wrote:
> > > The new keywords were the problem last time. To get the preprocessor
output,
> > > use the -e -l switches in combination and look at string1.lst.
> > >
> > > -Walter
> > >
> > Have done this.... nothing looks out of place though the listing has
rather less in it then I had expected.
> >
> > However I have found that if I replace the use of '__i' with say 'i__i'
the problem goes away. Guess :>
> >
> > Seems odd to me and the listing does not show any use of '__i'.
> >
> > I have never liked using '__' in front of variables. But in this case
not much can be done.
> >
> > I will do a straight forward hack and replace '__i' with 'i__i' for now.
> >
> > If I have time I will try a small program without STL to see if I can
isolate the problem.
> >
> > Regards,
> > Damian
> >
> > >
> > > Damian Dixon wrote in message <1104_984729166@dilbert>...
> > > >On Thu, 15 Mar 2001 10:17:41 -0800, "Walter" <walter@digitalmars.com>
> > > wrote:
> > > >> What's the compiler error message you get?
> > > >>
> > > >>
> > > >
> > > >sc
> > >
regression\string1.cpp -mn -C -WA -S -3 -a8 -c -gf -D_CONSOLE=1 -Ab -Id:\sc\
> > > stltemp\STL\port\stlport -oregression\string1.obj
> > > >Error: d:\sc\stltemp\STL\port\stlport\stl/_iterator_base.h(387): ')'
> > > expected
> > > >Error: d:\sc\stltemp\STL\port\stlport\stl/_iterator_base.h(393): ')'
> > > expected
> > > >Error: d:\sc\stltemp\STL\port\stlport\stl/_iterator_base.h(405): ')'
> > > expected
> > > >Error: d:\sc\stltemp\STL\port\stlport\stl/_iterator_base.h(406):
> > > template-argument '_Distance' not used in function parameter types
> > > >Fatal Error:
d:\sc\stltemp\STL\port\stlport\stl/_iterator_base.h(414): too
> > > many errors
> > > >Lines Processed: 2729  Errors: 5  Warnings: 0
> > > >Build failed
> > > >
> > > >
> > > >This is the regression test in the STLport archive
'port/test/console.prj'
> > > .. No changes to the source or project settings other then setting the include path.
> > > >
> > > >I'm still looking at the problem, though time is rather short as I am
> > > running a week late on final delivery of a large project.
> > > >
> > > >I was going to look to see if any of the new keywords introduced were
> > > causing a name clash first, the next step was to look at preprocessor output. One of the problems with most portable
> > > >versions of STL is the use of macro's to change behaviour of the
method's,
> > > templates etc... to suit a particular compiler. This can make it very difficult to figure out exactly what is going on,
> > > >requireing the pre-processor output to be examined.
> > > >
> > > >I'll probably have some time on Sunday to have a look more deeply.
> > > >
> > > >Regards,
> > > >Damian
> > > >
> > > >> Damian Dixon wrote in message <1103_984661516@dilbert>...
> > > >> >Hi,
> > > >> >
> > > >> >I've just downloaded the new compiler and have tried to compile
the
> > > >> regression tests for STLport 4.0 (the version currently in cvs).
> > > >> >
> > > >> >I'm sorry to report that the tests did not compile.
> > > >> >
> > > >> >I will try to figure out what may be causing the problem and
update the
> > > >> version in cvs as soon as I can.
> > > >> >
> > > >> >In the mean time if anyone has any fixes could you either post
them or
> > > >> email them to me if you don't have access as a commiter to cvs.
> > > >> >
> > > >> >Regards
> > > >> >Damian
> > > >> >
> > > >>
> > > >>
> > > >
> > > >
> > >
> > >
> >
> >
>
>

www.ansi.org

you can buy the C99 and C++ standard as electronic form (pdf) each for $18

better then paying $200 for the hardcopy :>

The search on the site is not too great as it turns up the hardcopy and not the softcopy. If you have problems finding the softcopy standards let me know and I will email the page.

Jan Knepper <jan@smartsoft.cc> wrote in message news:3AB29704.5A62F4C8@smartsoft.cc...
> I would like to have that document as well
> I guess is also covers C++?!
> Where did you get it?
>
> Don't worry, be Kneppie!
> Jan
>
>
>
> Damian Dixon wrote:
>
> > Oh, I've also obtained the ANSI C99 document as a reference, at $18 I thought it was the cheapest way of getting a definitive reference :>
>
>
>

Walter <walter@digitalmars.com> wrote in message news:98ueld$lu2$1@digitaldaemon.com...
> Regardless of the __i keyword problem, the internal error is a compiler
bug.
> Could you zip up a test case and email it to me please? walter@digitalmars.com
>
>
I thought that was the case. I will try to zip up a case. The problem is I
may not be able to
reduce the test case to a small self contained test :<.

The problem occurs when buildig the regression tests, after correcting the '__i's.

I  think the easiest thing is if I update STLport via CVS first, and let you know when I have updated the code.

I will of course try and produce a reduced test case as well.

Any hints as to what the compiler is having problems with.

Damian
>
> "Damian Dixon" <damian.dixon@tenetdefence.com> wrote in message news:1105_984746733@dilbert...
> > I've now run into a full stop..... I've changed the '__i's to 'i__i' and
> now:
> >
> > sc
>
console.cpp -mn -C -WA -S -3 -a8 -c -gf -D_CONSOLE=1 -Ab -Id:\sc\stltemp\STL
> \port\stlport -oconsole.obj
> > sc
>
regression\string1.cpp -mn -C -WA -S -3 -a8 -c -gf -D_CONSOLE=1 -Ab -Id:\sc\
> stltemp\STL\port\stlport -oregression\string1.obj
> > Fatal Error: D:\SC\stltemp\STL\port\test\regression\string1.cpp(32):
> internal error cg87 1144
> > Lines Processed: 16360  Errors: 1  Warnings: 0
> > Build failed
> >
> > I will have to look at a small stand alone test :<
> >
> > Any suggestions would be welcome.
> >
> > On Fri, 16 Mar 2001 12:40:49 GMT, Damian Dixon
> <damian.dixon@tenetdefence.com> wrote:
> > > On Fri, 16 Mar 2001 01:12:44 -0800, "Walter" <walter@digitalmars.com>
> wrote:
> > > > The new keywords were the problem last time. To get the preprocessor
> output,
> > > > use the -e -l switches in combination and look at string1.lst.
> > > >
> > > > -Walter
> > > >
> > > Have done this.... nothing looks out of place though the listing has
> rather less in it then I had expected.
> > >
> > > However I have found that if I replace the use of '__i' with say
'i__i'
> the problem goes away. Guess :>
> > >
> > > Seems odd to me and the listing does not show any use of '__i'.
> > >
> > > I have never liked using '__' in front of variables. But in this case
> not much can be done.
> > >
> > > I will do a straight forward hack and replace '__i' with 'i__i' for
now.
> > >
> > > If I have time I will try a small program without STL to see if I can
> isolate the problem.
> > >
> > > Regards,
> > > Damian
> > >
> > > >
> > > > Damian Dixon wrote in message <1104_984729166@dilbert>...
> > > > >On Thu, 15 Mar 2001 10:17:41 -0800, "Walter"
<walter@digitalmars.com>
> > > > wrote:
> > > > >> What's the compiler error message you get?
> > > > >>
> > > > >>
> > > > >
> > > > >sc
> > > >
>
regression\string1.cpp -mn -C -WA -S -3 -a8 -c -gf -D_CONSOLE=1 -Ab -Id:\sc\
> > > > stltemp\STL\port\stlport -oregression\string1.obj
> > > > >Error: d:\sc\stltemp\STL\port\stlport\stl/_iterator_base.h(387):
')'
> > > > expected
> > > > >Error: d:\sc\stltemp\STL\port\stlport\stl/_iterator_base.h(393):
')'
> > > > expected
> > > > >Error: d:\sc\stltemp\STL\port\stlport\stl/_iterator_base.h(405):
')'
> > > > expected
> > > > >Error: d:\sc\stltemp\STL\port\stlport\stl/_iterator_base.h(406):
> > > > template-argument '_Distance' not used in function parameter types
> > > > >Fatal Error:
> d:\sc\stltemp\STL\port\stlport\stl/_iterator_base.h(414): too
> > > > many errors
> > > > >Lines Processed: 2729  Errors: 5  Warnings: 0
> > > > >Build failed
> > > > >
> > > > >
> > > > >This is the regression test in the STLport archive
> 'port/test/console.prj'
> > > > .. No changes to the source or project settings other then setting
the
> > > > include path.
> > > > >
> > > > >I'm still looking at the problem, though time is rather short as I
am
> > > > running a week late on final delivery of a large project.
> > > > >
> > > > >I was going to look to see if any of the new keywords introduced
were
> > > > causing a name clash first, the next step was to look at
preprocessor
> > > > output. One of the problems with most portable
> > > > >versions of STL is the use of macro's to change behaviour of the
> method's,
> > > > templates etc... to suit a particular compiler. This can make it
very
> > > > difficult to figure out exactly what is going on,
> > > > >requireing the pre-processor output to be examined.
> > > > >
> > > > >I'll probably have some time on Sunday to have a look more deeply.
> > > > >
> > > > >Regards,
> > > > >Damian
> > > > >
> > > > >> Damian Dixon wrote in message <1103_984661516@dilbert>...
> > > > >> >Hi,
> > > > >> >
> > > > >> >I've just downloaded the new compiler and have tried to compile
> the
> > > > >> regression tests for STLport 4.0 (the version currently in cvs).
> > > > >> >
> > > > >> >I'm sorry to report that the tests did not compile.
> > > > >> >
> > > > >> >I will try to figure out what may be causing the problem and
> update the
> > > > >> version in cvs as soon as I can.
> > > > >> >
> > > > >> >In the mean time if anyone has any fixes could you either post
> them or
> > > > >> email them to me if you don't have access as a commiter to cvs.
> > > > >> >
> > > > >> >Regards
> > > > >> >Damian
> > > > >> >
> > > > >>
> > > > >>
> > > > >
> > > > >
> > > >
> > > >
> > >
> > >
> >
> >
>
>

Walter <walter@digitalmars.com> wrote in message news:98ueip$llf$1@digitaldaemon.com...
> __i is a new keyword added to the compiler. It is there to support
imaginary
> numbers, and is the imaginary number 1, and is evidently the main source
of
> the difficulties you are having.
>

yep....

> Keywords starting with __ and with _ followed by an uppercase letter are reserved for the implementation. STL seems to think that __ identifiers
are
> reserved to itself :-(.
>

I know, sadly enough not every one follows the guidelines. I think it is a carry over from the fact that SGI wrote most of the original implementation for thier compiler and made the work available to others.

Hence why I ran the regression tests to see what would happen.

If you want to let me know by email when a new version of the compiler is out and I will look at running STL port through the regression tests ASAP.

I tend to sometimes have too many distractions to always check the newsgroup :< for announcments.

> Does the STL port use complex.h? If so, it should change to using oldcomplex.h.

Probably. I need to check. STL port may also provide its own verison of complex. If it does then the header will more then likly be names 'complex' rather then 'complex.h'

Damian
>
> -Walter
>
>
> "Damian Dixon" <damian.dixon@tenetdefence.com> wrote in message news:1104_984746449@dilbert...
> > On Fri, 16 Mar 2001 01:12:44 -0800, "Walter" <walter@digitalmars.com>
> wrote:
> > > The new keywords were the problem last time. To get the preprocessor
> output,
> > > use the -e -l switches in combination and look at string1.lst.
> > >
> > > -Walter
> > >
> > Have done this.... nothing looks out of place though the listing has
> rather less in it then I had expected.
> >
> > However I have found that if I replace the use of '__i' with say 'i__i'
> the problem goes away. Guess :>
> >
> > Seems odd to me and the listing does not show any use of '__i'.
> >
> > I have never liked using '__' in front of variables. But in this case
not
> much can be done.
> >
> > I will do a straight forward hack and replace '__i' with 'i__i' for now.
> >
> > If I have time I will try a small program without STL to see if I can
> isolate the problem.
> >
> > Regards,
> > Damian
> >
> > >
> > > Damian Dixon wrote in message <1104_984729166@dilbert>...
> > > >On Thu, 15 Mar 2001 10:17:41 -0800, "Walter" <walter@digitalmars.com>
> > > wrote:
> > > >> What's the compiler error message you get?
> > > >>
> > > >>
> > > >
> > > >sc
> > >
>
regression\string1.cpp -mn -C -WA -S -3 -a8 -c -gf -D_CONSOLE=1 -Ab -Id:\sc\
> > > stltemp\STL\port\stlport -oregression\string1.obj
> > > >Error: d:\sc\stltemp\STL\port\stlport\stl/_iterator_base.h(387): ')'
> > > expected
> > > >Error: d:\sc\stltemp\STL\port\stlport\stl/_iterator_base.h(393): ')'
> > > expected
> > > >Error: d:\sc\stltemp\STL\port\stlport\stl/_iterator_base.h(405): ')'
> > > expected
> > > >Error: d:\sc\stltemp\STL\port\stlport\stl/_iterator_base.h(406):
> > > template-argument '_Distance' not used in function parameter types
> > > >Fatal Error:
d:\sc\stltemp\STL\port\stlport\stl/_iterator_base.h(414):
> too
> > > many errors
> > > >Lines Processed: 2729  Errors: 5  Warnings: 0
> > > >Build failed
> > > >
> > > >
> > > >This is the regression test in the STLport archive
> 'port/test/console.prj'
> > > .. No changes to the source or project settings other then setting the include path.
> > > >
> > > >I'm still looking at the problem, though time is rather short as I am
> > > running a week late on final delivery of a large project.
> > > >
> > > >I was going to look to see if any of the new keywords introduced were
> > > causing a name clash first, the next step was to look at preprocessor output. One of the problems with most portable
> > > >versions of STL is the use of macro's to change behaviour of the
> method's,
> > > templates etc... to suit a particular compiler. This can make it very difficult to figure out exactly what is going on,
> > > >requireing the pre-processor output to be examined.
> > > >
> > > >I'll probably have some time on Sunday to have a look more deeply.
> > > >
> > > >Regards,
> > > >Damian
> > > >
> > > >> Damian Dixon wrote in message <1103_984661516@dilbert>...
> > > >> >Hi,
> > > >> >
> > > >> >I've just downloaded the new compiler and have tried to compile
the
> > > >> regression tests for STLport 4.0 (the version currently in cvs).
> > > >> >
> > > >> >I'm sorry to report that the tests did not compile.
> > > >> >
> > > >> >I will try to figure out what may be causing the problem and
update
> the
> > > >> version in cvs as soon as I can.
> > > >> >
> > > >> >In the mean time if anyone has any fixes could you either post
them
> or
> > > >> email them to me if you don't have access as a commiter to cvs.
> > > >> >
> > > >> >Regards
> > > >> >Damian
> > > >> >
> > > >>
> > > >>
> > > >
> > > >
> > >
> > >
> >
> >
>
>

Damian Dixon wrote in message <98v5ql$11lp$2@digitaldaemon.com>...
>Any hints as to what the compiler is having problems with.


It's happening in the complex number code. But stl shouldn't have any complex numbers in it!

I have found & fixed a couple problems in the complex number support. If you want, I can email the test compiler to you, to see if the internal error still appears.

In general, I still think there must be an __i in the STL source there somewhere that's screwing things up.

-Walter


Damian Dixon wrote in message <98v6dh$1264$1@digitaldaemon.com>...
>Walter <walter@digitalmars.com> wrote in message news:98ueip$llf$1@digitaldaemon.com...
>> __i is a new keyword added to the compiler. It is there to support
>imaginary
>> numbers, and is the imaginary number 1, and is evidently the main source
>of
>> the difficulties you are having.
>>
>
>yep....
>
>> Keywords starting with __ and with _ followed by an uppercase letter are reserved for the implementation. STL seems to think that __ identifiers
>are
>> reserved to itself :-(.
>>
>
>I know, sadly enough not every one follows the guidelines. I think it is a carry over from the fact that SGI wrote most of the original implementation for thier compiler and made the work available to others.
>
>Hence why I ran the regression tests to see what would happen.
>
>If you want to let me know by email when a new version of the compiler is out and I will look at running STL port through the regression tests ASAP.
>
>I tend to sometimes have too many distractions to always check the newsgroup :< for announcments.
>
>> Does the STL port use complex.h? If so, it should change to using oldcomplex.h.
>
>Probably. I need to check. STL port may also provide its own verison of complex. If it does then the header will more then likly be names 'complex' rather then 'complex.h'
>
>Damian
>>
>> -Walter
>>
>>
>> "Damian Dixon" <damian.dixon@tenetdefence.com> wrote in message news:1104_984746449@dilbert...
>> > On Fri, 16 Mar 2001 01:12:44 -0800, "Walter" <walter@digitalmars.com>
>> wrote:
>> > > The new keywords were the problem last time. To get the preprocessor
>> output,
>> > > use the -e -l switches in combination and look at string1.lst.
>> > >
>> > > -Walter
>> > >
>> > Have done this.... nothing looks out of place though the listing has
>> rather less in it then I had expected.
>> >
>> > However I have found that if I replace the use of '__i' with say 'i__i'
>> the problem goes away. Guess :>
>> >
>> > Seems odd to me and the listing does not show any use of '__i'.
>> >
>> > I have never liked using '__' in front of variables. But in this case
>not
>> much can be done.
>> >
>> > I will do a straight forward hack and replace '__i' with 'i__i' for
now.
>> >
>> > If I have time I will try a small program without STL to see if I can
>> isolate the problem.
>> >
>> > Regards,
>> > Damian
>> >
>> > >
>> > > Damian Dixon wrote in message <1104_984729166@dilbert>...
>> > > >On Thu, 15 Mar 2001 10:17:41 -0800, "Walter"
<walter@digitalmars.com>
>> > > wrote:
>> > > >> What's the compiler error message you get?
>> > > >>
>> > > >>
>> > > >
>> > > >sc
>> > >
>>
>regression\string1.cpp -mn -C -WA -S -3 -a8 -c -gf -D_CONSOLE=1 -Ab -Id:\sc
\
>> > > stltemp\STL\port\stlport -oregression\string1.obj
>> > > >Error: d:\sc\stltemp\STL\port\stlport\stl/_iterator_base.h(387): ')'
>> > > expected
>> > > >Error: d:\sc\stltemp\STL\port\stlport\stl/_iterator_base.h(393): ')'
>> > > expected
>> > > >Error: d:\sc\stltemp\STL\port\stlport\stl/_iterator_base.h(405): ')'
>> > > expected
>> > > >Error: d:\sc\stltemp\STL\port\stlport\stl/_iterator_base.h(406):
>> > > template-argument '_Distance' not used in function parameter types
>> > > >Fatal Error:
>d:\sc\stltemp\STL\port\stlport\stl/_iterator_base.h(414):
>> too
>> > > many errors
>> > > >Lines Processed: 2729  Errors: 5  Warnings: 0
>> > > >Build failed
>> > > >
>> > > >
>> > > >This is the regression test in the STLport archive
>> 'port/test/console.prj'
>> > > .. No changes to the source or project settings other then setting
the
>> > > include path.
>> > > >
>> > > >I'm still looking at the problem, though time is rather short as I
am
>> > > running a week late on final delivery of a large project.
>> > > >
>> > > >I was going to look to see if any of the new keywords introduced
were
>> > > causing a name clash first, the next step was to look at preprocessor output. One of the problems with most portable
>> > > >versions of STL is the use of macro's to change behaviour of the
>> method's,
>> > > templates etc... to suit a particular compiler. This can make it very difficult to figure out exactly what is going on,
>> > > >requireing the pre-processor output to be examined.
>> > > >
>> > > >I'll probably have some time on Sunday to have a look more deeply.
>> > > >
>> > > >Regards,
>> > > >Damian
>> > > >
>> > > >> Damian Dixon wrote in message <1103_984661516@dilbert>...
>> > > >> >Hi,
>> > > >> >
>> > > >> >I've just downloaded the new compiler and have tried to compile
>the
>> > > >> regression tests for STLport 4.0 (the version currently in cvs).
>> > > >> >
>> > > >> >I'm sorry to report that the tests did not compile.
>> > > >> >
>> > > >> >I will try to figure out what may be causing the problem and
>update
>> the
>> > > >> version in cvs as soon as I can.
>> > > >> >
>> > > >> >In the mean time if anyone has any fixes could you either post
>them
>> or
>> > > >> email them to me if you don't have access as a commiter to cvs.
>> > > >> >
>> > > >> >Regards
>> > > >> >Damian
>> > > >> >
>> > > >>
>> > > >>
>> > > >
>> > > >
>> > >
>> > >
>> >
>> >
>>
>>
>
>

On Sat, 17 Mar 2001 00:09:35 -0800, "Walter" <walter@digitalmars.com> wrote:
> 
> Damian Dixon wrote in message <98v5ql$11lp$2@digitaldaemon.com>...
> >Any hints as to what the compiler is having problems with.
> 
> 
> It's happening in the complex number code. But stl shouldn't have any complex numbers in it!
> 
> 
ok.. I will have a look to see if I missed anything to do with '__i'.

Thanks
Damian

Jan,

Not sure if I followed up on this....

The C++ and C99 electronic copies of the standards can be obtained for $18 each from:

http://webstore.ansi.org/ansidocstore/default.asp

And select the department:
   NCITS (ITI): Information Technology (includes X3, NCITS, ANSI/ISO)

You actually have to select the item hitting Go did not work for me.

You will find the links for buying the documents near the bottom.

Not a very easy web-site to navigate.

Regards,
Damian

On Fri, 16 Mar 2001 17:43:16 -0500, Jan Knepper <jan@smartsoft.cc> wrote:
> I would like to have that document as well
> I guess is also covers C++?!
> Where did you get it?
> 
> Don't worry, be Kneppie!
> Jan
> 
> 
> 
> Damian Dixon wrote:
> 
> > Oh, I've also obtained the ANSI C99 document as a reference, at $18 I thought it was the cheapest way of getting a definitive reference :>
> 
> 
>