Re: [9front] 7c: compile error after sysupdate

["Ethan Gardener" <eekee57@fastmail.fm>]

On Mon, Jun 24, 2019, at 8:05 PM, cinap_lenrek@felloff.net wrote:
> > would it be a good idea to have sysupdate source rootstub?
> 
> i worked on this for a day and then gave up. its too hard and
> too complicated.
> 
> sysupdate would need some check to update itself in case
> there needs to be stuff like this. but how do you deploy
> THAT change? the only way is to fork everything. have
> a thumbstone commit with the sysupdate change. and then
> the next sysupdate goes to a new repository.

hmm...  i've got an idea.  as you say below, it won't help with
adding dirs to /, but just for fun:

i don't think you'd need a new repository, but you would need a way
for sysupdate to check and update itself first.  the first update
would be trouble, you're right there.  after that, it wouldn't be
complex with one or two helper scripts.  sysupdate could have just 2
lines (plus #!), the first running sysupdate-updater, the second execs
sysupdate-doit.  sysupdate-updater updates sysupdate-doit, but
sysupdate and sysupdate-updater would be updated by hg as normal,
because they're not running when hg is.  not that the 2-line sysupdate
would ever need updating.

i could code this, if you'd like it.

> on the other hand, its not so easy to automate creating
> stuff in the root. initially, the root of the fileserver
> was readonly and the only way to create directories there
> was console command. what if you don't run sysupdate on
> the fileserver but from a cpu server? then you can't
> access the fileserver console.

oh yes, of course.