From mboxrd@z Thu Jan 1 00:00:00 1970 From: John Court To: Daniel Jacobowitz Cc: Mark Kettenis , gdb-patches@sources.redhat.com Subject: Re: [RFA] The MIPS/Linux port Date: Thu, 05 Jul 2001 17:45:00 -0000 Message-id: <3B4509F6.40901@oz.agile.tv> References: <20010704121952.A17445@nevyn.them.org> <20010704163423.A25013@nevyn.them.org> <3B44FEE6.2070506@oz.agile.tv> <20010705173013.A18881@nevyn.them.org> X-SW-Source: 2001-07/msg00135.html Daniel Jacobowitz wrote: >On Fri, Jul 06, 2001 at 09:57:26AM +1000, John Court wrote: > >>Daniel Jacobowitz wrote: >> >>>On Wed, Jul 04, 2001 at 11:51:27PM +0200, Mark Kettenis wrote: >>> >>>>Looks like the MIPS port was started before a bit of reformatting of >>>>the sources and conversion to ISO C took place. Please >>>> >>> >>> >>>Fixed. Do you see any other formatting problems? >>> >>Hi Daniel, >> >>You have a bug in your supply_fpregset() and fill_fpregset() >>routines. You are being passed an elf_fpregset_t * and incrementing >>the counter by 264 od bytes each time through the loops. You >>need to offload the arguement something like the following before >>using it with regi. >> >> elf_fpreg_t *fpregp = &(*fpregsetp)[0]; >> >> >>This wont show up very well until you try and load a core dump >>files which will actually cause gdb to die as it spins off into invalid >>memory. I would be surprised if your floating point regs in >>gdb look any good the way the code is at the moment. >>I had to look at the assembler to work this out a few days ago. >> > >No, you're wrong. Note that I say: > supply_register (FCRCS_REGNUM, (char *)(*fpregsetp + 32)); > >&(*fpregsetp)[0] is just fancy notation for *fpregsetp, in this case at >least. > >I specifcally fixed that bug at least three times :) > > In this case I am happy to be wrong :-) I was working with someone elses patches on 5.0 and missed the extra "*" in your core. I should have checked more closely.....I have now and you are right :-(. Cheers John