#10 C++ parser can't parse std C++ header

[Stefan Seefeld]

invoking 'make' in demo/C++ yields

../../synopsis -c config.py
-Wc,parser=C++,formatter=HTML,linker=Linker -o html
GapBuffer.hh
/usr/include/g++-3/stl_alloc.h:344: parse error before
`_Obj *'
/usr/include/g++-3/stl_alloc.h:351: parse error before
`void *'
/usr/include/g++-3/stl_alloc.h:398: parse error before
`: static'
/usr/include/g++-3/stl_alloc.h:406: parse error before
`( __n'
/usr/include/g++-3/stl_alloc.h:409: parse error before
`__my_free_list ='
/usr/include/g++-3/stl_alloc.h:417: parse error before
`* __my_free_list'
/usr/include/g++-3/stl_alloc.h:418: parse error before
`( __result'
/usr/include/g++-3/stl_alloc.h:420: parse error before
`__r ;'
/usr/include/g++-3/stl_alloc.h:422: parse error before
`* __my_free_list'
/usr/include/g++-3/stl_alloc.h:423: parse error before
`( __result'
errors while parsing file /tmp/synopsis-x0o66k

no clue what the problem is at this point (i.e. a real
parse error, wrong context, i.e. incomplete macro
expansion, etc.). Thus it makes no sense to attach the
file in question...

[Nobody/Anonymous]

Logged In: NO

The problem lies in the fact that C++ comment are not
removed eg:
line 344 of stl_alloc is:
static _Obj* __VOLATILE _S_free_list[_NFREELISTS];
which is expanded to:
static _Obj* volatile // Needed at -O3 on SGI
_S_free_list[_NFREELISTS];
because in std_alloc.h we have a line 114:
# define __VOLATILE volatile // Needed at -O3 on SGI
< this
above is not removed>

[Stephen Davies]

Logged In: YES
user_id=768

OCC handles comments just fine, so that shouldn't be a problem.

Only the first parse error has any macro anyway.

I just tried compiling it with the debug "occ.gdb", and it
works fine, once I figured out you have to parse alloc.h to
get both stl_config.h first and then stl_alloc.h.

[Nobody/Anonymous]

Logged In: NO

The point I have tried to made is:
static _Obj* volatile // Needed at -O3 on SGI
_S_free_list[_NFREELISTS];
every things on the same line will not be compiled correctly
whatever
is the compiler.
I do not know why using occ.gdb makes it work, I could only
guess that
somehow the comment are discarded.
--Fathi

[Stephen Davies]

Logged In: YES
user_id=768

Okay, I see what you're saying. The form screwed up your
text.. you're saying that everything is on the same line,
with the comment in the middle?

When I parse it here, I don't get a comment. What OS are you
using? From the comment it sounds like SGI.

I wrote up a little test case that #defines a macro with a
comment like that, and the comment it is not included
whether I use gcc's cpp or the ucpp used by Synopsis. I also
tried running Synopsis on the file and got no parse errors.

Can you try with the latest version from CVS?

[Nobody/Anonymous]

Logged In: NO

Sorry I can't get it from CVS.
I cannot recall on wich OS I reproduced the problem
but it can only be Solaris 2.5 - 2.8 or Linux RH7.0.
Since I am unable to compile Synopsis-0.4.1 under
Solaris I think it might have been Linux. I'll give it a try
at home.
The comment about SGI only acknowledge that
the implementation of STL is a derived one from SGI.
Which remind me to state that in both case the compiler
I use is gcc.
--Fathi

[Stefan Seefeld]

Logged In: YES
user_id=764

hmm, it has been a while, but I remember it to be a bug in a
RH specific gcc version. The bug was precisely that the
preprocessor, when called with option '-C' correctly kept
comments, but expanded macros incorrectly (i.e. removed
newlines, leading to '//' comments in the middle of a line
of code...