1/1 
|
Building issues for eCos toolchain using latest CVS
by Unknown on Mar 2, 2004 |
Not available! | ||
|
I have been building the toolchain needed to build and debug applications
with eCos (I am only using the minimal configuration) and have encountered a
couple of issues.
I am using CVS sources checked out this morning (about 1100 GMT) of
binutils, gcc-3.2.3 and gdb-5.3
1) Heiko Panther's patch (see
http://www.opencores.org/forums.cgi/openrisc/2003/11/00024) needs to be
applied to gcc-3.2.3 to build correctly according to the eCos toolchain
building guidelines (see http://ecos.sourceware.org/build-toolchain.html).
2) The file gdb-5.3/gdb/or1k-tdep.c has been checked in with carriage
returns in which is not ideal for Linux environment - I would suggest it is
cleaned up using something like 'perl -pe 's/\r\n/\n/' -i or1k-tdep.c'
first. It should not really reside in the CVS repository like this.
3) The file gdb-5.3/gdb/or1k-tdep.c does not build on my machine (SuSE 7.2,
running GCC 2.95.3). A minor change to this fixed it:
Index: or1k-tdep.c
===================================================================
RCS file: /cvsroot/anonymous/or1k/gdb-5.3/gdb/or1k-tdep.c,v
retrieving revision 1.4
diff -u -r1.4 or1k-tdep.c
--- or1k-tdep.c 16 Feb 2004 13:50:25 -0000 1.4
+++ or1k-tdep.c 2 Mar 2004 17:53:28 -0000
@@ -895,6 +895,8 @@
function prologue.. */
unsigned long insn = read_memory_integer(func_pc+4,4);
+ int code, r1, r2, idx_h, idx;
+
/* If this is not the l.sw I(r1),r2 we're looking for,
give it a second try. */
if( (insn&0xFC1FF800) != 0xd4011000 )
@@ -906,11 +908,11 @@
then our frame pointer lies at the offset specified
by bits [25-21][10-0]. */
- int code = insn >> 26;
- int r1 = (insn & 0x001F0000) >> 16;
- int r2 = (insn & 0x0000F800) >> 11;
- int idx_h = (insn & 0x03E00000) >> 10;
- int idx = (insn & 0x000007FF) | idx_h;
+ code = insn >> 26;
+ r1 = (insn & 0x001F0000) >> 16;
+ r2 = (insn & 0x0000F800) >> 11;
+ idx_h = (insn & 0x03E00000) >> 10;
+ idx = (insn & 0x000007FF) | idx_h;
if(code == 0x35 && r1 == 1 && r2 == 2)
{
Robert Cragie, Design Engineer
_______________________________________________________________
Jennic Ltd, Furnival Street, Sheffield, S1 4QT, UK
http://www.jennic.com Tel: +44 (0) 114 281 2655
_______________________________________________________________
|
|||
|
Building issues for eCos toolchain using latest CVS
by Unknown on Mar 3, 2004 |
Not available! | ||
|
Hi Robert,
you probably noticed that there is no single person maintainer. Instead we use collaborative approach to maintain the tools. Developers that don't seem to act like beginner software develoeprs and show experience dealing with complex SW projects and team based projects, get access to the cvs. Since you are working on the latest version, and your fixed your local copy, the easiest way is to commit your changes back to the cvs. If you don't have access, please let me know what is your opencores username and I'll add you to maintainers. regards, Damjan
1) Heiko Panther's patch (see
http://www.opencores.org/forums.cgi/openrisc/2003/11/00024) needs to be
applied to gcc-3.2.3 to build correctly according to the eCos toolchain
building guidelines (see http://ecos.sourceware.org/build-toolchain.html).
2) The file gdb-5.3/gdb/or1k-tdep.c has been checked in with carriage
returns in which is not ideal for Linux environment - I would suggest it
is
cleaned up using something like 'perl -pe 's/\r\n/\n/' -i or1k-tdep.c'
first. It should not really reside in the CVS repository like this. 3) The file gdb-5.3/gdb/or1k-tdep.c does not build on my machine (SuSE 7.2,
running GCC 2.95.3). A minor change to this fixed it:
|
|||
1/1