[Crash-utility] [PATCH 0/5] let make flexible for multiple host

[Dave Anderson]

----- Original Message -----
> (2012/04/17 23:44), Dave Anderson wrote:
> > 
> > 
> > ----- Original Message -----
> >> Hi Dave,
> >>
> >> This patch set will make configure or build flexible for host environment,
> >> in particular dealing with cross compiler.
> >> Please consider about this proposal.
> >>
> >> patch#1:
> >>   By declaring tables about target and host relationships,
> >>   cleanup corresponding code and the new build combination will be
> >>   adapted briefly.
> >>
> >> Remaining patches:
> >>   Try to relieve cross compiler burden by using explicit
> >>   host_cross option.
> >>   I'm using cross compiler environ host=x86 for target=ppc,
> >>   and might attempt target=<another arch>  in the near future.
> >>   I stocked private patches to accomplish cross compiling
> >>   but I'd like to discard them permanently with your agreement.
> >>
> >> Thanks,
> >> Toshi
> > 
> > I am presuming that you have successfully built a PPC binary on an
> > x86 or x86_64 host, and then you have copied the binary to a PPC host
> > that doesn't have a resident compiler?  But then again, looking at
> > your
> 
> Hi Dave,
> 
> I didn't explain about background issues.
> 
> I built a PPC binary on an x86 host cross compiler. When run it,
> I have to ftp to PPC machine where dose not have self build tools.
> Here is dumping my build environment info.
>  $ uname -m
>  x86_64
>  $ file gcc
>  gcc: ELF 32-bit LSB executable, Intel 80386, version 1 (SYSV),
>  dynamically linked (uses shared libs), for GNU/Linux 2.2.5, stripped
>  $ ./gcc -dumpmachine
>  powerpc-linux-gnu
> 
> > patch as is, it looks like you can only allow "target=PPC" on a
> > PPC64 host,
> > which is already supported?:
> > 
> >     static struct host_table {
> > 
> >     ... [ cut ] ...
> > 
> >     }, ppc[] = {
> >             {
> >                     .host = PPC64,
> >                     .cflags = "-m32 -D_FILE_OFFSET_BITS=64 -fPIC",
> >                     .gdb_conf_flags = "--target=ppc-elf-linux
> >                     CFLAGS=-m32",
> >             }
> >     };
> 
> This "target=PPC" on a PPC64 host has been already supported
> since I copied it into ppc[] table.
>  [Crash-utility] [PATCH v1 1/4] Support PPC32 Core analysis on PPC64 Host
>   https://www.redhat.com/archives/crash-utility/2011-November/msg00019.html
> 
> The purpose is to build ppc 32bit binary with compatible target CFLAGS "-m32"
> of ppc64 compiler. This is similar idea with x86_64.
> 
> But I added new cross support(is_cross is TRUE) in patch#5 which is
> distinct from original purpose.
> 
> +	{
> +		.host = X86,
> +		.cflags = "-D_FILE_OFFSET_BITS=64",
> +		.gdb_conf_flags = "--host=i686-pc-linux-gnu "
> +				  "--target=powerpc-elf-linux",
> +		.is_cross = TRUE,
> +	},
> 
> > In any case, I would prefer that you specifically address your
> > particular PPC-on-X86 build issue, because I'm not sure what
> > other combinations would be particularly useful.
> 
> > I'm not particularly interested in speculative changes.  The only
> > future architecture support I can see on the horizon is 64-bit ARM,
> > and that is a few a few years away.  And we are lucky that x86 and
> > ARM have the same endianness and data-type sizes, so we can continue
> > (hopefully) to build x86_64 binaries to analyze 64-bit ARM vmcores
> > on x86_64 hosts.
> 
> My issues are to arrange original crash package as compilable one with cross.
> I edit package's Makefile or configure parameters --host/--target at spec file,
> finally building with environment variables CC, AR, LD, etc.
> 
> By applying patches, I can build crash only with tiny spec update,
> "make target=PPC host_cross=X86" without modification of released
> archive.
> 
> Then, I'll start mips architecture (seems to go toward embedded minority).
> There are probably the similar considerations about cross compile.
> I want to be done with "make target=MIPS host_cross=X86" by adding
> mips tables.

Wait a minute -- how about putting the horse before the cart?  There
is a significant amount of work involved with supporting MIPS as an 
architecture, which would have to be accomplished first.  

> 
> > Also, note that your patch fails because it collides with Rabin's queued
> > fix for building SIAL for target=ARM and target=PPC builds:
> > 
> >   [Crash-utility] [PATCH] build sial with TARGET_CFLAGS
> >   https://www.redhat.com/archives/crash-utility/2012-March/msg00113.html
> 
> I haven't caught this update. I'd send updated portion.
> 
> Thanks,
> Toshi

My hope was that you could accomplish the cross-compile automatically
without invoking the extra "host_cross=X86" option support.  If
configure sees that you're trying a "target=PPC" on an x86 or x86_64,
then it could presume it's your unique cross-compile scenario, and
then do whatever it has to do from that point on.

I also was under the impression that a MIPS binary was not all that
beneficial to the embedded folks, because it is not a typical 
situation where the crash analysis could take place *on* the embedded
MIPS platform.

Dave