Suivez-nous sur X
|
|
|
0,
A,
B,
C,
D,
E,
F,
G,
H,
I,
J,
K,
L,
M,
N,
O,
P,
Q,
R,
S,
T,
U,
V,
W,
X,
Y,
Z,
ALL
|
|
0,
A,
B,
C,
D,
E,
F,
G,
H,
I,
J,
K,
L,
M,
N,
O,
P,
Q,
R,
S,
T,
U,
V,
W,
X,
Y,
Z
|
|
0,
A,
B,
C,
D,
E,
F,
G,
H,
I,
J,
K,
L,
M,
N,
O,
P,
Q,
R,
S,
T,
U,
V,
W,
X,
Y,
Z
|
|
A propos d'Obligement
|
|
David Brunet
|
|
|
|
Interview with Jason McMullan
(Interview conducted by Olivier Tigréat - August 2010)
|
|
Here an interview with Jason McMullan, an american developper who works since some times on the improvement of AROS, specially
in the 68k port front.
 Would you please tell us about
yourself and how you became interested in computers?
I live in the United States, and was born in 1974. At an early age, I
became interested in electronic, and after I found analog and digital
electronics too difficult for my 8 year old brain to handle, I became
interested in computer programming.
Although most schools at the time used the Apple IIe computer, my
first computer was a TI 99/4A (due to Texas Instruments dumping them
onto the market at $50 per machine!). From there, it was a rapid
progression to the IBM PC models (via the Tandy 1000EX) and the BBS
scene, DOS x86 assembly, C, Pascal, and then to a boarding school
(with a VAX 11/780), and college (Carnegie Mellon University), where I
was exposed to functional languages, Linux, and all sort of other
interesting things.
From there, I've primarily been an Embedded Systems Engineer, working
on device drivers and low level operating system interfaces, primary
using Linux as my daily operating system of choice.
And how did you become interested
in AROS?
A few years ago, I got the computer I always wanted as a kid - an
Amiga 1000. In the process of trying to maximize its functionality
last year, I wanted to be able to use Linux to cross-compile programs
to run on my A1000.
So I found AROS, which had all the sources I need, but no m68k-amiga
port. So I took up the challenge, and started work on it.
What computers do you own these
days? And what do you use them for?
For AROS, I have an AMD beige box for development (running Linux, of
course), an Amiga 1000, Amiga 2000, Amiga 1200, and (most) of the
parts for an Amiga 4000.
Laying around I also have a DEC VT101, Apple Mac Quadra (68k),
PowerComputing PowerPC Mac clone, Apple iMac (green CRT model), various
IBM PC pieces and parts.
How would you summarize your
work on AROS so far?
I've primarily been involved in the amiga-m68k port, but have dabbled
in lots of the common code in AROS, including the lightweight
Workbench replacement (Workbook), the DOS Packets work (initial BCPL
support, and the ABI IoFS->DOS Packets conversion), boot sequence
work, work on reducing the AROS memory footprint... Wherever I have an
itch, I scratch it.
AROS Bootstrap
Workbook
However, I must say that without Toni Wilen's help, assistance, and
handholding, we would not have achieved the KickStart Replacement
bounties so quickly. I'm more of a systems and compilers guy, but Toni
knows all the ins and outs of Amiga hardware, and almost all of the
strange ROM quirks of the various KickStarts and peripherial ROMs.
You've started to implement your
light replacement for WorkBench, called Workbook (as Wanderer is too big to be used in basic 68k offers). And by the way now Mazze
is porting Scalos... What are your thoughts about the situation on this front, and what are your plans for Workbook?
Workbook will stay as a minimal in-ROM Workbench replacement (needed
for AmigaOS Workbench floppies and some games), but I don't plan on
adding too many features at the moment.
Scalos will be a fine 'fuller featured' Workbench for Amiga m68k, and
I expect Wanderer to remain as the 'fullest featured' Workbench,
unless Scalos surpasses it.
All three currently have their own niche in the AROS ecosystem, but I
do expect that users will settle on either Wanderer or Scalos as the
'preferred' Workbench in the next few years.
If I understood correctly, you
are currently also working on SATA drives, and especially on their hotplug ability... Can you tell us more about it?
I am working on AHCI SATA support, and to do so I have needed to add
some infrastructure to AROS to support that. The first major change
was to more fully support ejectable media in the AROS boot sequence,
and that was recently completed, and checked in. Pavel Fedin has been
instrumental in removing the last few bugs in the code on the x86_64
and i386 platforms.
The next change, which I hope to push before the end of August, is a
set of routines which help device drivers monitor and manage RDB
filesystems, including automatically dismounting filesystems when
ejected.
It's functionally complete, and is on a branch in my gitorious.org GIT
repo, but I'm not quite happy with its API quite yet. I'm currently
looking at the MorphOS mount.library API to get some inspiration.
After that, I'll push the AHCI driver (which is skeletal at the
moment, awaiting these previous changes). I expect that it should be
fully debugged by mid September.
Those modern drives will need
modern file systems. In your opinion, can SFS (and PFS which was recently open-sourced and ported) be
improved enough to handle nowadays' needs?
Yes, SFS and PFS should be sufficient. SFS has a problem with
dismounting on partitioned removable media (part of the changes I'm
working on at the moment), but other than that both of these should be
fine for the near future.
Especially on the side of
the partition size which is still limited to few hundreds of GB.
I'm not aware of the specific details of the SFS and PFS
implementations, so I can't really comment on that.
MorphOS just got a sort of a
"FUSE.library"... And what about the dos.library? Will it be easy to improve it to handle real big files,
such as a complete ISO image of a whole DVD? You have made a huge work on switching AROS to the use of dos-packets:
will it be of any use on this topic?
The AROS DOS Packets implementation is being tested on x86_64 (64-bit)
architectures as it's being developed, and we are removing 32-bit
restrictions as we find them.
Although accessing files larger than 4GB will probably not be supported
on 32-bit AROS, due to the substantial API changes that would be
involved, 64-bit AROS should end up with 64-bit safe file access by
the end of the ABIv1 update cycle.
Could you tell us about your
other plans for AROS in the future?
- More hardware support. I am working on getting an A4000 system up
with a Mediator PCI, so that I can extend expansion.library on AROS
with PCI bus functionality, and support some more common PCI devices.
- A amiga-m68k install set, which would consist of an ADF and a CDROM
image, that can be used to boot and run AROS on any supported Amiga
with Kickstart 1.3 or better.
- Reducing the AROS m68k RAM memory footprint, to be able to support
machines with 2MB of total RAM or less.
- Reducing the AROS m68k ROM memory footprint. An "AROS Lite" ROM
should be possible, with a OCS/ECS/AGA specific graphics.library and
input.library, that could fit into a 512K ROM.
What do you think about other
architectures for AROS, can it grow on them, and is ARM-port interesting?
I think the ARM and other architectures could be viable ports, **IF** people open-source their AROS applications.
Asking each AROS developer (especially the single developer shops) to make and test builds for x86, x86_64, m68k, PowerPC,
and ARM is a large burden for them. I feel that the biggest successes of Linux is due to the application developer philosophy.
If someone wants a port, they can just recompile your sources (And hopefully contribute theirchanges back to you!).
What is your opinion about
bounties? Are they mandatory to boost the development?
I don't feel that bounties are mandatory, but they do obviously help. I know that I wouldn't have worked so hard on the
initial m68k portwithout the Kickstart I bounty!
What are your thoughts about
the history of AROS? Things have considerably evolved since some years, but do you like the (historical) AROS moto: "No Schedule
And Rocking"?
I completely ignore the history of AROS. I try to keep politics out of
my development. I have no grudge against MorphOS or AmigaOS 4, nor any
fanatic fever that AROS is the best.
AROS just happens to have been the one that let me see its source
code, so it's the one I develop on. It's that simple.
Is AROS a nice place
to have fun for a programmer?
In general, yes. Although the time zone difference (I am in GMT-5)
means that I am often wake up to some surprising commits in the
morning. But in general, everyone is pretty friendly. And while, like
any development team, we occasionally have some mailing list
arguments, I think that we can usually 'agree to disagree' instead of
getting angry with each other.
I have the feeling (but may be
wrong...) that the dev-ml is only the emerged part of the iceberg, about your relations with some other devs
(especially Toni Wilen, but also Pavel Fedin or Staf Verhaegen for example): right? If so, could you tell us more about those
co-operations?
Pavel finds Skype to be a better communication medium for his style,
so I usually confer with him in my morning/his evening about AROS
development details. Toni Wilen and I often exchange emails about m68k
and Amiga specific questions that the list may not be interested in
(things like BCPL emulation quirks, UAE/WinUAE differences, etc.).
Do you use AROS yourself?
If not so, do you plan to at some point?
Nope. I'm a Linux guy. Although I find developing for AROS to be a fun
recreational programming exercise, I feel more comfortable in a
POSIX/Unix environment.
What would you like to see
improved in AROS?
Compatibility with AmigaOS 1.3 - AmigaOS 3.9 on the m68k architecture.
That's all I really care about.
Anything you would like to say?
Although it can be argued that I am not the most ardent AROS
supporter, nor plan on using it personally, I really enjoy working on
it - it suits my 'embedded systems' mindset. It's an interesting
puzzle box, one that I hope I never finish unlocking.
I've learned a lot about the 'Library OS' concept (which, oddly
enough, Microsoft seems to be developing an interest in lately with
MinWin), and some of the advantages and disadvantages of the APIs that
AmigaOS developed.
Which are?
Some of the AmigaOS advantages I like are things like:
- No system call overhead.
- Message passing by reference (instead of copying the entire messagepacket via a pipe or FIFO).
- User-level interrupt handlers.
But these advantages are solely possible by the biggest disadvantage of AmigaOS and AmigaOS derived operating systems -
minimal to nonexistent memory protection, and a single-CPU architecture. Pavel Fedin is working on better MMU and SMP support
for AROS, and I wish him the best of luck. The underlying AmigaOS architecture makethis a very interesting problem.
Not to mention learning m68k assembly, which I had not been exposed to before last fall (2010).
Also, I would like to say that MorphOS and AmigaOS could benefit from
opening up their source code - people such as myself are often looking
for an interesting challenge to their programming skills, and
releasing some of their kernel and library sources could help develop
more interest (and ports to other hardware!) of their systems.
|
