From mboxrd@z Thu Jan 1 00:00:00 1970 From: Daniel Jacobowitz To: Andrew Cagney Cc: Eli Zaretskii , gdb-patches@sources.redhat.com Subject: Re: [RFA] mips32_next_pc overdue cleanup Date: Thu, 05 Jul 2001 13:34:00 -0000 Message-id: <20010705133346.A15705@nevyn.them.org> References: <3B44CE13.8070405@cygnus.com> X-SW-Source: 2001-07/msg00112.html On Thu, Jul 05, 2001 at 04:29:07PM -0400, Andrew Cagney wrote: > Eli wrote: > > > According to the masks in the opcodes > >> directory, these changes are rigt, and what we had before can't possibly > >> work. Especially the fact that we would never, ever predict a BNE as taken > >> before these changes convinced me that they really were correct and this > >> code had never been particularly tested. Comments? > > > > > > I'd appreciate a short test case for this. I'd like to verify that > > the previous stuff indeed didn't work and that the new stuff does. > > > > Assuming that this code is run on SGI Irix, that is. > > > I'm wondering the same thing. > > > I'll try to make one. I'm not immediately sure how. > > > > It looks almost as if this code was for an unreleased port inside > > Cygnus - anyone know if that's true? I see active work on it in 1998 > > and 2000 in the ChangeLogs, and no mention of removing calls to it > > - but nothing uses it at all. > > > I'll do some digging. Thanks. I'm curious. Is there a CVS repository backed up somewhere which predates 1999, when it seems the sourceware copy was created? Makes it a little more challenging to do the archaelogy. > One thing, it is very easy for a software single step function to be > implemented yet never tested - everything supports h/w singlestep > now-a-days. Everything except my target :) The thing is, the (trivial) software single step function wasn't even in mips-tdep.c. Only all the complicated support routines for it were. The lossage isn't all that surprising, given that these functions had no callers. -- Daniel Jacobowitz Carnegie Mellon University MontaVista Software Debian GNU/Linux Developer