84930 - nsIRequest.idl:165: Error

Status: VERIFIED INVALID

(Core :: XPCOM, defect, P3)

Description

[horton]

yManagerComponent.idl nsICachingChannel.idl ../../../dist/idl

nsIRequest.idl

../../../dist/bin/xpidl -m header -w -I ../../../dist/idl -I. -o _xpidlgen/nsIRequest nsIRequest.idl

nsIRequest.idl:165: Error: unsigned const declaration 'VALIDATE_NEVER' initialized with negative constant

gmake[4]: *** [_xpidlgen/nsIRequest.h] Error 1

gmake[4]: Leaving directory `/usr/src/mozilla/netwerk/base/public'

gmake[3]: *** [export] Error 2

gmake[3]: Leaving directory `/usr/src/mozilla/netwerk/base'

gmake[2]: *** [export] Error 2

gmake[2]: Leaving directory `/usr/src/mozilla/netwerk'

gmake[1]: *** [export] Error 2

gmake[1]: Leaving directory `/usr/src/mozilla'

gmake: *** [build] Error 2



This is me trying to build mozilla on a slackware linux 7.1 box with the latest gnome 1.4 and X 4.1.0

Comment 1

[horton]

I also cannot build Netscape6 or use the latest mozilla builds.
I get this error if I try.

[horton@creep:~/mozilla-installer] sudo ./mozilla-installer
./mozilla-installer: line 48:  1812 Segmentation fault      ./mozilla-installer-bin --sync

Here is me stracing mozilla-installer-bin

open("/tmp/.tmp.xi.1/xpcom.xpi", O_RDONLY) = 6
open("/tmp/.tmp.xi.1/bin/libplds4.so", O_WRONLY|O_CREAT|O_TRUNC, 0666) = 7
open("/tmp/.tmp.xi.1/bin/libxpcom.so", O_WRONLY|O_CREAT|O_TRUNC, 0666) = 7
open("/tmp/.tmp.xi.1/bin/libmozz.so", O_WRONLY|O_CREAT|O_TRUNC, 0666) = 7
open("/tmp/.tmp.xi.1/bin/components/libjar50.so", O_WRONLY|O_CREAT|O_TRUNC, 0666) = 7
open("/tmp/.tmp.xi.1/bin/components/libunicharutil.so", O_WRONLY|O_CREAT|O_TRUNC, 0666) = 7
open("/tmp/.tmp.xi.1/bin/libmozjs.so", O_WRONLY|O_CREAT|O_TRUNC, 0666) = 7
open("/tmp/.tmp.xi.1/bin/libnspr4.so", O_WRONLY|O_CREAT|O_TRUNC, 0666) = 7
open("/tmp/.tmp.xi.1/install.js", O_WRONLY|O_CREAT|O_TRUNC, 0666) = 7
open("/tmp/.tmp.xi.1/bin/components/libxpinstall.so", O_WRONLY|O_CREAT|O_TRUNC, 0666) = 7
open("/tmp/.tmp.xi.1/bin/libxpistub.so", O_WRONLY|O_CREAT|O_TRUNC, 0666) = 7
open("/tmp/.tmp.xi.1/bin/libjsdom.so", O_WRONLY|O_CREAT|O_TRUNC, 0666) = 7
open("/tmp/.tmp.xi.1/bin/libplc4.so", O_WRONLY|O_CREAT|O_TRUNC, 0666) = 7
open("/libxpistub.so", O_RDONLY)        = -1 ENOENT (No such file or directory)
--- SIGSEGV (Segmentation fault) ---
+++ killed by SIGSEGV +++

Comment 2

[cls]

Need more details.  Namely:

What version of mozilla are you trying to compile (0.9.1, cvs (include date of
pull))
What version of libIDL (libIDL-config --version).

Please attach a full build log including configure output as well as config.log
& config/autoconf.mk.

The installer problem is a separate issue and you should file a separate bug on it.

Comment 3

[horton]

[horton@creep:~] libIDL-config --version
0.6.8

Comment 4

[cls]

I don't know the details of a slackware 7.1 installation but I've never seen
this problem with RH6.2 + ORBit 0.5.0 (which includes libIDL 0.6.8). 
Reassigning to xpidl.

Comment 5

[David Bradley]

The line in question for the IDL is:

const unsigned long VALIDATE_NEVER            = 1 << 12;

Which 1 << 12 = 4096 which is what I have in my header.

The only thing I can think to suggest is to replace the shifted values with 
their results and see if you can build and report back. I'm looking at the error 
in the code now.

This wouldn't be on a 64 bit system would it?

Comment 6

[Keyser Sose]

Marking NEW.

Comment 7

[David Bradley]

Does the reporter have the current libIDL?

We haven't seen any problems with this on any of internal boxes, so my guess is
it's a libIDL version issue.

If that is not the problem then the only solution would be compute these
constants and put them in as numeric values. I believe jband add these not too
long ago. Before I do that, I'd like to have a report back that manually doing
this does solve the problem.

Comment 8

[David Bradley]

Is this still a problem?

I think the second comment is a different problem, it's faulting because it
can't open a file. Horton if you're still having this problem I'd suggesting
filing a separate bug, if one does not already exist for it.

Comment 9

[David Bradley]

Shaun Savage wrote me the following:

There was a thread about not compiling on RH7.1 using upgraded gcc-3.0


I have traced it through and have found out that during the parsing of the
expression
const unsigned long NAME_FOO  1 << 1

when the binop_eval is exectuted the (SHFTOP,a,b)  "a" and "b" point to NAME_FOO
node.
I would expect a=1 and b=1 but they point to the address of the string NAME_FOO

Not being a bison guru I would assume that there is a bug in bison. The MOZ is
perfect [;-)]
I am looking at
const_dcl_def :    TOK_CONST const_type new_ident '=' const_expr  {


const_expr : or_expr
.
.
.
shift_expr :    add_expr
|   shift_expr SHFTOP add_expr  {do_binop($$,IDL_SHFT,$1,$3);}


$1 and $3 point to new_ident

I guess that the problem is in the new_ident area

Comment 10

[David Bradley]

*** Bug 100200 has been marked as a duplicate of this bug. ***

Comment 11

[David Bradley]

This appears to be a Bison or possibly a compiler issue and not a Mozilla issue.
Is there any reason not to mark this invalid?

Comment 12

[David Bradley]

Marking this invalid since it is a Bison/compiler issue and not a Mozilla issue.

Comment 13

[Phil Schwartau]

Marking Verified -