Re: egcs-2.9.66 (B20.1) errors

Re: egcs-2.9.66 (B20.1) errors

[Phil Edwards]

On Tue, 6 Apr 1999 16:40:43    Mumit Khan wrote:
>
>The only time I've seen is when running GCC with -mno-cygwin mode or using
>include path that has mingw32 include files before g++ include files. The
>Mingw32 header directory contains alloc.h, which should be deleted (and
>the contents moved into malloc.h), which causes this.

That's exactly the situation here.  I am now beating
my head into pulp for not mentioning this:  the specs
file being used is the same one that I sent to the
list a few days earlier, asking for comments.  The
default mode is now "no-cygwin", and the mingw32
include path comes before all the others.  Again, I
apologize for not mentioning that.  Gee, ya think it
could make a difference...?

The difference in 'spces' is the only change.  Honest.


>Can you please send me more info?
>  
>  $ env > bug.txt
>  $ cygcheck -s -r >> bug.txt 2>&1 
>  $ gcc -c -v -save-temps foo.cc >> bug.txt 2>&1
>
>and email me bug.txt and foo.ii file.

I'll send them separately in a moment.  I should point
out that I'll generally be using gcc by itself from a
DOS box.  The only times I venture into bash are when
I need to build a GNU package, where I have to run
things like 'configure' et al.

Phil
It's gonna be one of /those/ weeks...



-----== Sent via Deja News, The Discussion Network ==-----
http://www.dejanews.com/  Easy access to 50,000+ discussion forums

--
Want to unsubscribe from this list?
Send a message to cygwin-unsubscribe@sourceware.cygnus.com

Re: egcs-2.9.66 (B20.1) errors

[Phil Edwards]

On Tue, 6 Apr 1999 16:40:43    Mumit Khan wrote:
>
>The only time I've seen is when running GCC with -mno-cygwin mode or using
>include path that has mingw32 include files before g++ include files. The
>Mingw32 header directory contains alloc.h, which should be deleted (and
>the contents moved into malloc.h), which causes this.

That's exactly the situation here.  I am now beating
my head into pulp for not mentioning this:  the specs
file being used is the same one that I sent to the
list a few days earlier, asking for comments.  The
default mode is now "no-cygwin", and the mingw32
include path comes before all the others.  Again, I
apologize for not mentioning that.  Gee, ya think it
could make a difference...?

The difference in 'spces' is the only change.  Honest.


>Can you please send me more info?
>  
>  $ env > bug.txt
>  $ cygcheck -s -r >> bug.txt 2>&1 
>  $ gcc -c -v -save-temps foo.cc >> bug.txt 2>&1
>
>and email me bug.txt and foo.ii file.

I'll send them separately in a moment.  I should point
out that I'll generally be using gcc by itself from a
DOS box.  The only times I venture into bash are when
I need to build a GNU package, where I have to run
things like 'configure' et al.

Phil
It's gonna be one of /those/ weeks...



-----== Sent via Deja News, The Discussion Network ==-----
http://www.dejanews.com/  Easy access to 50,000+ discussion forums

--
Want to unsubscribe from this list?
Send a message to cygwin-unsubscribe@sourceware.cygnus.com

egcs-2.9.66 (B20.1) errors

[Phil Edwards]

I've been experiencing two errors with the egcs-1.1.2
replacement for B20.1.  The first is a fatal error in
the compiler proper; I have no clue about the second.

The attached (too short?) text shows the error message.
The very first time I tried compiling, egcs hung at
the indicated point.  I broke with ^C, egcs printed
the rest of its error, and exited.  It never happened
again.

The error message itself, regarding the parser error,
is just damn confusing to me.  This same stuff builds
fine on a Unix system.  I don't even know how to begin
hunting this kind of bug down.  (Parameterized types
are neat, but cuss words I hate template debugging.)


The same happens in a minimal testcase.  Line 64-66
of bastring.h are:

   template <class charT, class traits = string_char_traits<charT>,
      class Allocator = alloc >
   class basic_string

It's either puking at the start of "class Allocator,"
or (more likely?) at the end of the traits bit on the
previous line.

Huh?




-----== Sent via Deja News, The Discussion Network ==-----
http://www.dejanews.com/  Easy access to 50,000+ discussion forums

   []>  make
   bison_pp -d -o LPParser.cc LPParser.y
   g++ -c -Wall -I. LPParser.cc
   In file included from D:/cygnus/cygwin-b20/H-i586-cygwin32/bin/../lib/gc
   c-lib/i586-cygwin32/egcs-2.91.66/../../../../../include/g++/string:6,
                    from lpparser.y:5:
   D:/cygnus/cygwin-b20/H-i586-cygwin32/bin/../lib/gcc-lib/i586-cygwin32/eg
   cs-2.91.66/../../../../../include/g++/std/bastring.h:65: parse error before `>'
[hangs here, breaks out on ^C]
   g++.c:2574: Internal compiler error in function execute

   []>

Re: egcs-2.9.66 (B20.1) errors

[Phil Edwards]

On Wed, 7 Apr 1999 22:22:20    Mumit Khan wrote:
>Hi Phil,
>
>I'd like to close the bug report, so please let me know if removing
>alloc.h improved your situation or not.
>
>Regards,
>Mumit

My apologies; I've been off-site since I submitted
the bug report and didn't think to check in remotely.
This cheesy DejaNews mail account isn't the easiest
client to work with.  :-)

Yes, removing (actually renaming, 'cause I'm a 
paranoid bastard) ...\H-i586-cygwin32\i586-cygwin32\
include\mingw32\alloc.h solved the <string> header
problem perfectly.  I'm following your example of
cc'ing this kind of thing to the list, to help out
our pal the search engine.

Since the cygwin list doesn't generate nearly as much
traffic as I had feared, I'm in the midst of moving
my subscruption from the digest to the piece-at-a-time
mode.  Hopefully I'll be able to report changes and
fixes faster in the future (I'm sure I'll break
something else).


Luck++;
Phil



-----== Sent via Deja News, The Discussion Network ==-----
http://www.dejanews.com/  Easy access to 50,000+ discussion forums

--
Want to unsubscribe from this list?
Send a message to cygwin-unsubscribe@sourceware.cygnus.com

Re: egcs-2.9.66 (B20.1) errors

[Mumit Khan]

On Tue, 6 Apr 1999, Phil Edwards wrote:

> 
> I've been experiencing two errors with the egcs-1.1.2
> 
> The same happens in a minimal testcase.  Line 64-66
> of bastring.h are:
> 
>    template <class charT, class traits = string_char_traits<charT>,
>       class Allocator = alloc >
>    class basic_string
> 
> It's either puking at the start of "class Allocator,"
> or (more likely?) at the end of the traits bit on the
> previous line.
> 

The only time I've seen is when running GCC with -mno-cygwin mode or using
include path that has mingw32 include files before g++ include files. The
Mingw32 header directory contains alloc.h, which should be deleted (and
the contents moved into malloc.h), which causes this.

Can you please send me more info?
  
  $ env > bug.txt
  $ cygcheck -s -r >> bug.txt 2>&1 
  $ gcc -c -v -save-temps foo.cc >> bug.txt 2>&1

and email me bug.txt and foo.ii file.

Regards,
Mumit



--
Want to unsubscribe from this list?
Send a message to cygwin-unsubscribe@sourceware.cygnus.com

Re: egcs-2.9.66 (B20.1) errors

[Phil Edwards]

On Wed, 7 Apr 1999 22:22:20    Mumit Khan wrote:
>Hi Phil,
>
>I'd like to close the bug report, so please let me know if removing
>alloc.h improved your situation or not.
>
>Regards,
>Mumit

My apologies; I've been off-site since I submitted
the bug report and didn't think to check in remotely.
This cheesy DejaNews mail account isn't the easiest
client to work with.  :-)

Yes, removing (actually renaming, 'cause I'm a 
paranoid bastard) ...\H-i586-cygwin32\i586-cygwin32\
include\mingw32\alloc.h solved the <string> header
problem perfectly.  I'm following your example of
cc'ing this kind of thing to the list, to help out
our pal the search engine.

Since the cygwin list doesn't generate nearly as much
traffic as I had feared, I'm in the midst of moving
my subscruption from the digest to the piece-at-a-time
mode.  Hopefully I'll be able to report changes and
fixes faster in the future (I'm sure I'll break
something else).


Luck++;
Phil



-----== Sent via Deja News, The Discussion Network ==-----
http://www.dejanews.com/  Easy access to 50,000+ discussion forums

--
Want to unsubscribe from this list?
Send a message to cygwin-unsubscribe@sourceware.cygnus.com

Re: egcs-2.9.66 (B20.1) errors

[Mumit Khan]

On Tue, 6 Apr 1999, Phil Edwards wrote:

> 
> I've been experiencing two errors with the egcs-1.1.2
> 
> The same happens in a minimal testcase.  Line 64-66
> of bastring.h are:
> 
>    template <class charT, class traits = string_char_traits<charT>,
>       class Allocator = alloc >
>    class basic_string
> 
> It's either puking at the start of "class Allocator,"
> or (more likely?) at the end of the traits bit on the
> previous line.
> 

The only time I've seen is when running GCC with -mno-cygwin mode or using
include path that has mingw32 include files before g++ include files. The
Mingw32 header directory contains alloc.h, which should be deleted (and
the contents moved into malloc.h), which causes this.

Can you please send me more info?
  
  $ env > bug.txt
  $ cygcheck -s -r >> bug.txt 2>&1 
  $ gcc -c -v -save-temps foo.cc >> bug.txt 2>&1

and email me bug.txt and foo.ii file.

Regards,
Mumit



--
Want to unsubscribe from this list?
Send a message to cygwin-unsubscribe@sourceware.cygnus.com

egcs-2.9.66 (B20.1) errors

[Phil Edwards]

I've been experiencing two errors with the egcs-1.1.2
replacement for B20.1.  The first is a fatal error in
the compiler proper; I have no clue about the second.

The attached (too short?) text shows the error message.
The very first time I tried compiling, egcs hung at
the indicated point.  I broke with ^C, egcs printed
the rest of its error, and exited.  It never happened
again.

The error message itself, regarding the parser error,
is just damn confusing to me.  This same stuff builds
fine on a Unix system.  I don't even know how to begin
hunting this kind of bug down.  (Parameterized types
are neat, but cuss words I hate template debugging.)


The same happens in a minimal testcase.  Line 64-66
of bastring.h are:

   template <class charT, class traits = string_char_traits<charT>,
      class Allocator = alloc >
   class basic_string

It's either puking at the start of "class Allocator,"
or (more likely?) at the end of the traits bit on the
previous line.

Huh?




-----== Sent via Deja News, The Discussion Network ==-----
http://www.dejanews.com/  Easy access to 50,000+ discussion forums

   []>  make
   bison_pp -d -o LPParser.cc LPParser.y
   g++ -c -Wall -I. LPParser.cc
   In file included from D:/cygnus/cygwin-b20/H-i586-cygwin32/bin/../lib/gc
   c-lib/i586-cygwin32/egcs-2.91.66/../../../../../include/g++/string:6,
                    from lpparser.y:5:
   D:/cygnus/cygwin-b20/H-i586-cygwin32/bin/../lib/gcc-lib/i586-cygwin32/eg
   cs-2.91.66/../../../../../include/g++/std/bastring.h:65: parse error before `>'
[hangs here, breaks out on ^C]
   g++.c:2574: Internal compiler error in function execute

   []>