| | [ Return to Bugs & Features | Roadmap 1.1 | SVN ⇄ GIT ]
STR #261
| Application: | FLTK Library |
|---|
| Status: | 1 - Closed w/Resolution |
|---|
| Priority: | 5 - Critical, e.g. nothing working at all |
|---|
| Scope: | 2 - Specific to an operating system |
|---|
| Subsystem: | Core Library |
|---|
| Summary: | 64-bit FLTK 1.1.4 compilation bus errors immediately on Solaris 8 |
|---|
| Version: | 1.1.4 |
|---|
| Created By: | johns.ks.uiuc |
|---|
| Assigned To: | mike |
|---|
| Fix Version: | 1.1.5rc1 |
|---|
| Update Notification: | |
|---|
Trouble Report Files:
Trouble Report Comments:
| |
| #1 | johns.ks.uiuc
08:15 Jan 26, 2004 |
| Building FLTK 1.1.4 in 64-bit mode on Solaris 8 yields an unusable build that crashes immediately with a bus error, both in my application as well as in all of the FLTK test apps. The configure script was run with these variables set:
setenv CC cc
setenv CFLAGS "-fast -xO3 -native -xarch=v9a"
setenv CXX CC
setenv CXXFLAGS "-fast -xO3 -native -xarch=v9a"
Also, the cursor test program fails to compile:
=== making test ===
Compiling cursor.cxx...
"cursor.cxx", line 40: Error: Cannot cast from void* to int.
1 Error(s) detected.
gmake[1]: *** [cursor.o] Error 1
Executing the test programs (or my app) results in a bus error
here (crash always occurs in the same place):
johns{896} dbx browser
(/opt/SUNWspro/bin/../WS6U1/bin/sparcv9/dbx) run
Running: browser
(process id 25017)
signal BUS (invalid address alignment) in _XData32 at 0xffffffff7ed5bbd8
0xffffffff7ed5bbd8: _XData32+0x0070: ldx [%i3], %o1
(/opt/SUNWspro/bin/../WS6U1/bin/sparcv9/dbx) where
=>[1] _XData32(0x10014e6b0, 0x0, 0x4, 0xffffffff7fffeebc, 0xb, 0x72), at 0xffffffff7ed5bbd8
[2] XChangeProperty(0x1, 0x1, 0xffffffff7fffeebc, 0x100157aec, 0x20, 0x0), at 0xffffffff7ed2cacc
[3] Fl_X::make_xid(0x0, 0x1, 0x147800, 0x147ae0, 0x100000000, 0x36), at 0x100023fd0
[4] Fl_Window::show(0x147800, 0x147ae0, 0x100000000, 0x64, 0x19, 0x100036836), at 0x10001f588
[5] main(0x10014a8c0, 0xffffffff7ffff638, 0xffffffff7ffff648, 0x0, 0x0, 0x100000000), at 0x10000def8
(/opt/SUNWspro/bin/../WS6U1/bin/sparcv9/dbx)
| |
| |
| #2 | johns.ks.uiuc
09:05 Jan 26, 2004 |
| The crash appears to be caused by the use of the "int" type where "long" should have been spedified. (after reading the man pages for XChangeProperty()...)
Changing "version" to be a long instead of an int prevents the crash.
I don't like the sizeof(int)*8 in the call to XChangeProperty(), as I believe that's also incorrect. The man page refers to those as a format specifier, not as a field size, and so "32" format specifier actually returns a long, which in a 64-bit build is not 32-bits, thus the source of the original problem.
(see line 1088 of Fl_x.cxx):
// Make it receptive to DnD:
long version = 4;
XChangeProperty(fl_display, xp->xid, fl_XdndAware,
XA_ATOM, sizeof(int)*8, 0, (unsigned char*)&version, 1);
| |
| |
| #3 | mike
17:43 Feb 25, 2004 |
| Fixed in CVS - the anonymous CVS repository will be updated at midnight PST. | |
[ Return to Bugs & Features ]
|
| |
