194701 - Build failure on Gcc 3.2.2

Status: VERIFIED WORKSFORME

(SeaMonkey :: Build Config, defect)

Description

[Vidar Hoel]

User-Agent:       Opera/6.11 (Linux 2.4.21-pre4 i686; U)  [en]
Build Identifier: Mozilla 1.3b

I use the following .mozconfig:

# Options for 'configure' (same as command-line options).
ac_add_options --with-prefix=/usr/local
ac_add_options --enable-default-toolkit=gtk2
ac_add_options --enable-toolkit-gtk2
ac_add_options --enable-svg
ac_add_options --with-default-mozilla-five-home=/usr/local/mozilla
ac_add_options --with-x
ac_add_options --with-system-zlib
ac_add_options --with-system-png
ac_add_options --with-system-jpeg
ac_add_options --enable-xft
ac_add_options --disable-mailnews
ac_add_options --enable-crypto
ac_add_options --enable-plaintext-editor-only
ac_add_options --disable-ldap
ac_add_options --disable-logging
ac_add_options --disable-accessibility
ac_add_options --disable-composer
ac_add_options --disable-installer
ac_add_options --disable-activex
ac_add_options --disable-tests
ac_add_options --disable-debug
ac_add_options --with-pthreads
ac_add_options --disable-dtd-debug
ac_add_options --enable-optimize="-O3 -mcpu=pentiumpro -march=pentiumpro"
ac_add_options --enable-java-supplement
ac_add_options --enable-strip


When I run a "gmake -f client.mk" this error appears:
<-cut->

make[4]: Entering directory `/usr/src/mozilla/string/src'
nsFragmentedString.cpp
c++ -o nsFragmentedString.o -c -DOSTYPE=\"Linux2.4\" -DOSARCH=\"Linux\" -DOJI -D_IMPL_NS_COM  -I../../dist/include/xpcom -I../../dist/include/string -I../../dist/include -I/usr/src/mozilla/dist/include/nspr      -I/usr/X11R6/include   -fPIC  -O3 -mcpu=pentiumpro -march=pentiumpro -mpreferred-stack-boundary=4  -I/usr/X11R6/include -fno-rtti -fno-exceptions -Wall -Wconversion -Wpointer-arith -Wcast-align -Woverloaded-virtual -Wsynth -Wno-ctor-dtor-privacy -pedantic -Wno-long-long -O3 -mcpu=pentiumpro -march=pentiumpro -mpreferred-stack-boundary=4 -fshort-wchar -pthread -pipe  -DNDEBUG -DTRIMMED -O3 -mcpu=pentiumpro -march=pentiumpro  -O3 -mcpu=pentiumpro -march=pentiumpro -mpreferred-stack-boundary=4  -I/usr/X11R6/include -DMOZILLA_CLIENT -include ../../mozilla-config.h -Wp,-MD,.deps/nsFragmentedString.pp nsFragmentedString.cpp
In file included from /usr/include/c++/3.2.2/new:41,
                 from ../../dist/include/string/nsBufferHandleUtils.h:36,
                 from ../../dist/include/string/nsSharedBufferList.h:47,
                 from ../../dist/include/string/nsFragmentedString.h:35,
                 from nsFragmentedString.cpp:26:
/usr/include/c++/3.2.2/cstddef:53: parse error before `unsigned'
In file included from ../../dist/include/string/nsBufferHandleUtils.h:36,
                 from ../../dist/include/string/nsSharedBufferList.h:47,
                 from ../../dist/include/string/nsFragmentedString.h:35,
                 from nsFragmentedString.cpp:26:
/usr/include/c++/3.2.2/new:51: parse error before `{' token
/usr/include/c++/3.2.2/new:56: destructors must be member functions
/usr/include/c++/3.2.2/new:56: virtual outside class declaration
/usr/include/c++/3.2.2/new:79: parse error before `unsigned'
/usr/include/c++/3.2.2/new:79: parse error before `::' token
/usr/include/c++/3.2.2/new:80: parse error before `unsigned'
/usr/include/c++/3.2.2/new:80: parse error before `::' token
/usr/include/c++/3.2.2/new:83: parse error before `unsigned'
/usr/include/c++/3.2.2/new:84: parse error before `unsigned'
/usr/include/c++/3.2.2/new:85: parse error before `&' token
/usr/include/c++/3.2.2/new:85: `operator delete' takes type `void*' as first 
   parameter
/usr/include/c++/3.2.2/new:86: parse error before `&' token
/usr/include/c++/3.2.2/new:86: `operator delete' takes type `void*' as first 
   parameter
/usr/include/c++/3.2.2/new:89: parse error before `unsigned'
/usr/include/c++/3.2.2/new: In function `void* operator new(unsigned int, 
   ...)':
/usr/include/c++/3.2.2/new:89: `__p' undeclared (first use this function)
/usr/include/c++/3.2.2/new:89: (Each undeclared identifier is reported only 
   once for each function it appears in.)
/usr/include/c++/3.2.2/new: At global scope:
/usr/include/c++/3.2.2/new:90: parse error before `unsigned'
/usr/include/c++/3.2.2/new:96: parse error before `}' token
In file included from ../../dist/include/string/nsFragmentedString.h:35,
                 from nsFragmentedString.cpp:26:
../../dist/include/string/nsSharedBufferList.h: In static member function 
   `static nsSharedBufferList::Buffer* 
   nsSharedBufferList::NewWrappingBuffer(PRUnichar*, PRUnichar*, PRUnichar*)':
../../dist/include/string/nsSharedBufferList.h:205: call of overloaded `
   operator new(unsigned int)' is ambiguous
<internal>:205: candidates are: void* operator new(unsigned int)
/usr/include/c++/3.2.2/new:89:                 void* operator new(unsigned int, 
   ...)
../../dist/include/string/nsBufferHandleUtils.h: In function `HandleT* 
   NS_AllocateContiguousHandleWithData(const HandleT*, unsigned int, const 
   StringT*) [with HandleT = nsSharedBufferList::Buffer, StringT = 
   nsDependentSingleFragmentSubstring]':
../../dist/include/string/nsBufferHandleUtils.h:183:   instantiated from `HandleT* NS_AllocateContiguousHandleWithData(const HandleT*, const StringT&, unsigned int) [with HandleT = nsSharedBufferList::Buffer, StringT = nsDependentSingleFragmentSubstring]'
../../dist/include/string/nsSharedBufferList.h:190:   instantiated from here
../../dist/include/string/nsBufferHandleUtils.h:152: call of overloaded `
   operator new(unsigned int)' is ambiguous
<internal>:152: candidates are: void* operator new(unsigned int)
/usr/include/c++/3.2.2/new:89:                 void* operator new(unsigned int, 
   ...)
../../dist/include/string/nsBufferHandleUtils.h: In function `HandleT* 
   NS_AllocateContiguousHandleWithData(const HandleT*, unsigned int, const 
   StringT*) [with HandleT = nsSharedBufferList::Buffer, StringT = nsAString]':
../../dist/include/string/nsBufferHandleUtils.h:183:   instantiated from `HandleT* NS_AllocateContiguousHandleWithData(const HandleT*, const StringT&, unsigned int) [with HandleT = nsSharedBufferList::Buffer, StringT = nsAString]'
../../dist/include/string/nsSharedBufferList.h:198:   instantiated from here
../../dist/include/string/nsBufferHandleUtils.h:152: call of overloaded `
   operator new(unsigned int)' is ambiguous
<internal>:152: candidates are: void* operator new(unsigned int)
/usr/include/c++/3.2.2/new:89:                 void* operator new(unsigned int, 
   ...)
make[4]: *** [nsFragmentedString.o] Error 1
make[4]: Leaving directory `/usr/src/mozilla/string/src'
make[3]: *** [libs] Error 2
make[3]: Leaving directory `/usr/src/mozilla/string'
make[2]: *** [tier_2] Error 2
make[2]: Leaving directory `/usr/src/mozilla'
make[1]: *** [default] Error 2
make[1]: Leaving directory `/usr/src/mozilla'
make: *** [build] Error 2

I assume this is not a problem with GCC, since I have built a complete system using it :-)
As always, just say if you need more info.

Reproducible: Always

Steps to Reproduce:

1.
2.
3.




Running Linux 2.4.20 and 2.4.21pre4
GCC 3.2.2
glibc 3.2.1

Comment 1

[David Baron :dbaron: (⌚️UTC-5, no longer working on Mozilla)]

Does this compiler setup compile other C++ programs correctly?  The errors are
in the C++ library header files, not ours.

Comment 2

[Mitch]

Remember also, compiling with gcc 3.x will preclude you from using 
Sun's java plugin with mozilla.

Comment 3

[hacker formerly known as [e-mail]]

Which is why you should switch to using the Blackdown version which has been
compiled with gcc 3.2,
http://www.blackdown.org/java-linux/java2-status/jdk1.4-status.html .  (Not that
this has anything to do with this actual bug.)


This looks like a possible dupe of bug 191009.  Do you have CPP set in your
environment?  If so, to what?  Can you attach the full output of the ./configure
run and the config.log file?

Comment 4

[Vidar Hoel]

Attachment: Output of ./configure

./configure &> configure.log

Comment 5

[Vidar Hoel]

Attachment: config.log

Comment 6

[Vidar Hoel]

Just included the files as wished for.

I have compiled other C++ programs on this system, like MPlayer, without any errors.

And yes, it could seem like a dupe of bug 191009, since it is the exact same
line in "cstddef". But I do not have CPP in my environment. (I assume you mean
that I have not done any "export CPP=<something>" or "export CFLAGS=<someyhing>")

Comment 7

[hacker formerly known as [e-mail]]

Well, the cause of the error is the same: size_t is being seen as undefined by
autoconf and being redefined as unsigned.  I don't see anything in the log to
explain why that's happening though.  In fact, all of those type tests (mode_t,
off_t, pid_t) I would expect to pass on Linux.  


Try running this sample program though the preprocessor and grep the output for
size_t.

cat >dummy.c <<EOF
#include <sys/types.h>
#include <stdlib.h>
#include <stddef.h>
EOF

gcc -E - dummy.c | grep size_t

You may also want to check the output to see which stddef.h is being included.

Comment 8

[Bradley Baetz (:bbaetz)]

You may want to delete config.cache, esp if you updated your compiler.

'Brad' on teh main tbox page is gcc-3.2.3pre, so it definately builds...

Comment 9

[Vidar Hoel]

Attachment: gcc -E dummy.c | grep size_t

Comment 10

[Vidar Hoel]

I assume you meant "gcc -E dummy.c" not "gcc -E - dummy.c"
Anyway, I attached the output, and I can not see anything wrong. I got the same
output on a box where I do manage to compile Mozilla.

The stddef.h he uses is
"/usr/lib/gcc-lib/i686-pc-linux-gnu/3.2.2/include/stddef.h", and that seems
fairly ok.

And yes, I do a "make distclean" every time.

Comment 11

[hacker formerly known as [e-mail]]

Ok, here's a silly question, do you have egrep installed?  The actual commands
used by configure are:

cat > conftest.$ac_ext <<EOF
#line 5994 "configure"
#include "confdefs.h"
#include <sys/types.h>
#if STDC_HEADERS
#include <stdlib.h>
#include <stddef.h>
#endif
EOF
if (eval "$ac_cpp conftest.$ac_ext") 2>&5 |
  egrep "(^|[^a-zA-Z_0-9])size_t[^a-zA-Z_0-9]" >/dev/null 2>&1; then
  rm -rf conftest*
  ac_cv_type_size_t=yes
else
  rm -rf conftest*
  ac_cv_type_size_t=no
fi
rm -f conftest*

Comment 12

[Dan Mosedale (:dmosedale, :dmose)]

FWIW, I've built mozilla with gcc 3.2.2 on red hat linux 7.2, and not seen this
problem.

Comment 13

[hacker formerly known as [e-mail]]

Marking WFM based upon dmose's comments and the 3.2.2* tinderboxes.

Comment 14

[Tracy Walker [:tracy]]

verified

Comment 15

[Vidar Hoel]

I am still not able to compile Mozilla on my box. The question about if I have egrep installed, I have to answer yes. I tried to debug the configure script, and I saw that both my boxes (one does compile it, and the other does not) set this:

ac_cv_type_size_t=no

In fact, both boxes gives this:
checking for mode_t... no
checking for off_t... no
checking for pid_t... no
checking for size_t... no

I really hope you can check this it one more time, because I want to get to the bottom of this. A complete reinstall of the whole box would be a sad defeat :-(