|
From: | Mike Thomas |
Subject: | Re: [Gcl-devel] Windows gcc 3.4.0 rdata custom relocation |
Date: | Sat, 24 Jul 2004 17:06:23 +1000 |
User-agent: | Mozilla/5.0 (Windows; U; Windows NT 5.0; en-US; rv:1.4) Gecko/20030624 Netscape/7.1 (ax) |
Hi Camm. Camm Maguire wrote:
Greetings! Mike, this is absolutely fantastic! Congratulations. You are the next relocation expert :-).
Thanks. Relocation to a tropical island with water, food shelter and medical facilities is what I assume you mean.
Don't know about the aref stuff -- haven't checked the patch closely, but I already got: GCL (GNU Common Lisp) 2.7.0 CLtL1 Jun 24 2004 15:18:57 Source License: LGPL(gcl,gmp), GPL(unexec,bfd) Binary License: GPL due to GPL'ed components: (READLINE BFD UNEXEC) Modifications of this banner must retain notice of a compatible license Dedicated to the memory of W. Schelter Use (help) to get some basic information on how to use GCL.(setq a (make-array 10))#(NIL NIL NIL NIL NIL NIL NIL NIL NIL NIL)(aref a -1)Error: Array index -1 out of bounds for #(NIL NIL NIL ...) Fast links are on: do (si::use-fast-links nil) for debugging Error signalled by AREF. Broken at AREF. Type :H for Help.
Yes, that's what I wanted to happen, in line with our co-compilers LispWorks and Corman Common Lisp.
I closed the bug but no email notification came to me and presumably not to you either so I suppose it went the way of the alleged reconnection of our CVS notifications.
Do you have a test case showing the failure of the old algorithm?
Bug report 7807 : https://savannah.gnu.org/bugs/index.php?func=detailitem&item_id=7807
Am wondering if your fantastic breakthrough warrants a 2.6.4.
Conservatively speaking I don't believe so, partly because we have specified gcc 3.3.1 for Windows 2.6.x anyway, and also because I had to make some larger changes today as follows:
I forgot to try out the case with no --enable-debug which sure enough failed because the rdata section was not at the same index as when stabs sections are present - moved from 6 to 4.
So I spent the day rewriting to search through the section headers and to assign offsets/indices as required and, as noted earlier, to help clear the way for any future moves to build with MS Visual C. These changes appear to work, but I think we should test them for a while to ensure I haven't broken custom relocation on the other platforms. So far I've only tested Maxima too.
I also tried the brand new gcc MinGW32 3.4.1 release candidate which built and tested Maxima as required.
Cheers Mike Thomas.
[Prev in Thread] | Current Thread | [Next in Thread] |