Re: [Hampshire] Packaging help needed

2010/1/17 Stephen Pelc <stephen@???>:
>
> I would have said that I don't have the skill set for this - I'm
> not, and don't want to be, a Linux kernel guru. However, I have
> talked to a number of people who produce proprietary Linux apps
> in niche markeys, and they all agree that packaging is a real
> PITA. We may even consider writing a Linux install system for
> human beings. Interested people are welcome to contact me, but
> please don't send me philosophical rants about the evils of
> proprietary software. My flame-proof suit is good - I've been on
> standards committees for many years.
>

I am not about to disagree with you. Your task of having to make
packages for all the different Linux platforms is a PITA.
How are you attempting your packaging? I would build a virtual machine
for each of your target platforms and then you should be able to use
the same script for all of them, without any special "if" lines to do
different stuff if it is for RPM or .deb or 64bit etc.
Packaging of user land apps has very little to do with the Linux
kernel, it is more to do with getting the dependencies correct for the
various versions of user-land libs.
It is the "cross compiling" that complicates things considerably.
You can have ARM and MIPS virtual machines running on a x86 Linux.
I believe Google use an ARM VM on top of an x86 CPU Linux for all
their android development.

>
> There is a significant proportion of the compiler-writing
> community who believe that gcc has stifled compiler innovation.
> In the opinion of many it's an "adequate" compiler. The x86-32
> and x86-64 compilers are have huge commercial input, but are far
> from best-in-class. For other targets, e.g. ARM and Cortex,
> despite the funding of Codesourcery by ARM, the gcc compiler
> just isn't very good.
>
> The downside of gcc is that it has become dominant in certain
> sectors because it is free (beer). Dominance doesn't mean good.
> Open source doesn't necessarily mean good (quality).
>

gcc 4 output is very good for the x86 32bit and 64bit CPUs now that
SSA is being used. The gcc source code base is very good quality.
I do not know how good or not it is for ARM output, but it is
improving, particularly as Google are now very active on the ARM
platform.
ARM CPU instructions are very different from x86 instructions so I can
understand the difficulty in making a single compiler good for both.

I do a lot of software/hardware reverse engineering work and getting
my tools to work on both x86 binaries and ARM binaries requires
considerable work.

Kind Regards

James