Re: [PATCH] ld/ldlang.c: fatal error on architecture mismatch

[Ian Lance Taylor <iant at google dot com>]

Daniel Jacobowitz <drow@false.org> writes:

> On Thu, May 03, 2007 at 09:28:27AM -0700, Ian Lance Taylor wrote:
> > Daniel Jacobowitz <drow@false.org> writes:
> > 
> > > Would anyone be more amenable to making it an error with an entry in
> > > NEWS or in the manual describing how to build an image without an
> > > entry point?  Someone else will have to write it; I never do that.
> > 
> > Building an image without an entry point is trivial: link as usual,
> > then objcopy -O binary.
> 
> I don't know what you mean, but it's not what I mean.  That binary
> still has an entry point; it's just not recorded in the binary file.
> It may be in the ELF file, and tools which use the ELF file (like
> GDB, or static analysis tools) may use it.

Sure.  The entry point is implicit in the binary file.  It may be at
the start of the file, or it may be at a fixed address.  That works
fine today no matter what you use for ENTRY in the linker script, or
whether you use ENTRY at all.  You may get a linker warning which you
will always ignore.  Your proposal will change that to a hard error.

I'm not going to continue replying on this thread.  I think I've made
my point: the proposed change will break some currently working
builds.  This isn't hypothetical: I know builds that will break.  That
doesn't mean we can't do it, it's just something to consider.

Ian