~vdupras

Quebec, Canada

Troubadour des internets

~vdupras/collapseos

Last active 4 months ago
View more

Recent activity

Re: Goodbye Sourcehut 4 months ago

From Virgil Dupras to ~vdupras/collapseos

On Tue, Feb 2, 2021, at 10:01 AM, raingloom wrote:
> > Why do I use git? Is it for collaboration? No, because there is no
> > collaborator.
> Well, I was gonna get back to my ATA driver code, so this is a
> disappointment.
> I specifically didn't want to start yet another OS.
>

As you can read a few lines below in the same message, this is a
statement of fact, not of intention: you can verify that fact by looking
at the commit log of the project.

If you want to contribute to Collapse OS, you are welcome to.

Re: Broken DKIM header in mailing list 5 months ago

From Virgil Dupras to ~sircmpwn/sr.ht-discuss

On Tue, Jan 5, 2021, at 10:42 AM, Simon Ser wrote:
> Do you have the raw e-mails, before and after being mangled by lists.sr.ht?
> Can you upload them somewhere (without changing the line endings) and link
> them to me?

E-mails in my Sent folder don't include DKIM headers (they are added by
Fastmail), I'm not sure they can help.

However, since it appears that every mail I send triggers the bug, this very email
should contain the information you seek.

Regards,
Virgil

Goodbye Sourcehut 5 months ago

From Virgil Dupras to ~vdupras/collapseos

The email delivery problem outlined in the previous thread[1] and the
lack of response to my related inquiry[2] made me rethink my reasons for
leaving Github to Sourcehut[3]. Therein lies a problem:

1. SourceHut is free software and is managed as a public and open
   project.
2. I benefit from this project because it hosts my own project.
3. Usually, when encountering a problem within such a project, the
   ethical thing to do is to solve it yourself. It's not Drew's burden
   to fix this email problem.
4. However, I have no desire to delve into SourceHut's code to fix a
   SMTP bug (if that bug is indeed in SourceHut's code).
5. I don't want to leave this problem unsolved either.

Broken DKIM header in mailing list 5 months ago

From Virgil Dupras to ~sircmpwn/sr.ht-discuss

A problem that appears to be related to sourcehut was recently reported
on Collapse OS' mailing list[1]. It seems that some emails' DKIP headers
are incorrectly mangled when the mailing list sends its emails to sub-
scribers: It uses \r\r\n line endings instead of \r\n ones.

Off-list debugging took place and determined that:

1. I seem to be the only source of problematic emails (my email provider
   is Fastmail).
2. The problem persists even after I've fixed my own DKIM signature
  which was previously incorrect.
3. When I send emails directly, headers are fine.

Do you think that the problem could originate from a bug in Sourcehut?

Re: Dogfooding Collapse OS 5 months ago

From Virgil Dupras to ~vdupras/collapseos

Thanks Bradley for this additional information.

I've rechecked my email provider's configuration and noticed that my DKIM configuration was wrong. I've fixed it. To be honest, I didn't know anything about DKIM until yesterday.

To verify whether this mis-configuration triggered a bug in sourcehut, I'm sending this message and making sure that Michael is not in the CC field. If this was indeed the trigger, I think we have enough information to file a bug.

Dogfooding Collapse OS 5 months ago

From Virgil Dupras to ~vdupras/collapseos

I've been exclusively "dogfooding" Collapse OS in the last few days, that is,
improving it from within itself on my TRS-80 4P. Collapse OS is already pretty
usable! Becoming a Forth user is fun.

I've recently acquired a Tandy Coco2 and I'm aching to get into Motorola's
beautiful pond. I've been asking myself whether I'd go in with the extra dog-
fooding challenge. What if I developed and deployed the 6809 port of Collapse
OS entirely in Collapse OS, on a TRS-80 4P? That would be hard, but pretty
cool. I think I'll give it a try.

This would definitely mark a "dogfooding" turn into Collapse OS' development.
Up until now, I've been relying on emulation on a POSIX platform to build my
stuff, and then iron out the kinks on the real machine. Cozy, efficient.

Collapse OS self-hosts on a Sega Master System 6 months ago

From Virgil Dupras to ~vdupras/collapseos

A significant milestone was reached today: I assembled Collapse OS from
within itself on a Sega Master System. It had a dual EEPROM socket, a SD
card adapter as well as a PS/2 keyboard adapter. From the current blkfs
written to a SD card and with the help of the "~AT28" override, I could
execute SMS xcomp steps and end up with a brand new bootable Collapse OS
ROM!

The SMS is the tightest machine (8Kb of RAM) to assemble Collapse OS yet
and I'm not planning on targeting a tighter machine. This means that I'm
done finding the bottom of the barrel, specs-wise and focus on other
goals.

I'm planning on focusing on usability next. To that end, I'll pay myself
a trip down the vintage road and go meta, self-hosting-wise: develop

What I've been up to 6 months ago

From Virgil Dupras to ~vdupras/collapseos

This mailing list doesn't see a lot of activity. Given the low "approachability" of the project, I can understand why. To dispel the idea that discussion around this project is impossible, I'll begin posting status updates regularly here, at the risk of writing to cricket sounds.

In the last month, I've been working on improving support for the Sega Master System, with self-hosting as the end goal of this streak. 

One challenge was that VE was unusable in screens narrower than 64 columns. That was solved by making the contents in VE horizontally scrollable. In fact, there's been a whole slew of improvement with regards to Grid applications (what we would call curses applications on POSIX, except the Grid writes directly to video, it doesn't go through a TTY).

Another annoyance was the 32 column limit. I've been aware for a long while that the VDP supports a text mode which allows for 40 columns, but I hadn't got around yet to implement support for it. Now it's done and it makes Collapse OS much more usable.

Then, I wanted to access SD cards from it. My first idea was to build a bridge from the cartridge connector to a RC2014 connector, which would have allowed me to reuse my SPI relay prototype. It took a while for me to wire such a bridge and I should have read the SMS datasheet more carefully first: the I/O address space is already 100% used by the system. You can't use I/Os for something else than built-in peripherals without breaking the system. I fell back on a SPI relay that would use a controller port. It's slow, but it works!

This is pretty much where I am at now, self-hosting is almost within reach. One problem with the SMS is that 8K of memory is not enough to hole both the assembler and the would-be-assembled binary. This binary has to be written directly to EEPROM. Because assemblers already use the "Addressed devices" mechanism, it shouldn't be much of a challenge to "pipe" the output of the assembler to EEPROM. However, the only ROM holder I have for now is a single socket. I need to build a dual-socket ROM holder so that I can write the ROM to the second socket. Once this is done, Collapse OS should be self-hosting on the SMS.

Re: [PATCH] Replace "-ansi" with "-std=c89" in emul/Makefile 10 months ago

From Virgil Dupras to ~vdupras/collapseos

Applied, thanks.

Re: Application software development 10 months ago

From Virgil Dupras to ~vdupras/collapseos

On Wed, Jul 22, 2020 at 06:16:34AM -1000, Dustin Rawlings wrote:
> As I was reading through the project roadmap on the webpage, I noticed
> that there wasn't much by way of application software mentioned there.

Because the scope of Collapse OS is very narrow, I'm not planning on having a
very large set of applications in it. That being said, it doesn't mean that
there aren't worthwhile apps left to develop. For example, a hex editor would
be very handy. But no, there's nothing on the roadmap regarding applications.

> Something in that vein that I immediately wanted after loading up the
> cvm was some kind of pager for quickly scrolling through the blkfs
> blocks. It gets tedious typing "N LIST" over and over when I just want
> to step through the blocks in order for a while.