Use spec in lib/collections expression.

[jtobin]

4b2a3d8 replaced

  =/  c=[term cord]  ((hard ,[term cord]) b)

with

  =/  c=[term cord]  ;;([term cord] b)

but without the comma, [term cord] is not parsed as a spec. As a result, collections errors on boot. This commit replaces the cell with a pair, yielding a proper spec.

[belisarius222]

Seems like the real problem here is that ;; doesn't properly switch the parser into pattern mode when in flat form. The comma on the first sub-hoon shouldn't be necessary.

[ohAitch]

It having that parser mode is the reason to keep it despite being redundant with %- even!

…

On Wed, May 15, 2019 at 6:13 AM Ted Blackman ***@***.***> wrote: Seems like the real problem here is that ;; doesn't properly switch the parser into pattern mode when in flat form. The comma on the first sub-hoon shouldn't be necessary. — You are receiving this because you are subscribed to this thread. Reply to this email directly, view it on GitHub <#1171?email_source=notifications&email_token=AAOFPBXPOTUTAZFQT3FNXCLPVQD7BA5CNFSM4HNCF332YY3PNVWWK3TUL52HS4DFVREXG43VMVBW63LNMVXHJKTDN5WW2ZLOORPWSZGODVOTSAA#issuecomment-492648704>, or mute the thread <https://github.com/notifications/unsubscribe-auth/AAOFPBTBOUWXHEXPT2WJTCDPVQD7BANCNFSM4HNCF33Q> .

[joemfb]

As of #1164, the %mcmc parser on master does use pattern mode . @jtobin, I'm not seeing this collections failure on boot, how can I reproduce it?

[joemfb]

I suspect you're booting into a mismatch between compiler and userspace - a compiler that doesn't have %mcmc changes and a userspace that expects them. And I further suspect that this is happening in the new cc-release pill staging scripts. The fake-ship provisioning was recently simplified to make CI work better, but it makes these mismatches more likely. Long story short, the pill update scripts need to boot ships without -A. This is my bad ....

[pilfer-pandex]

Yeah, the new ;; should parse its p (first subexpression) as a spec; if this is not happening, it should be fixed in the parser, rather than at use site like in this change, but I'm pretty sure I made that change to the parser. If joe's right, this is caused by having an old version of the compiler in a pill; if so, it's good we're discovering this now. Accordingly, I'm rejecting.

As for pair, the only difference between pair and a [] with two elems, as far as I am aware, is the addition of p and q faces. It would really be nice if we didn't have to have faces to do ordinary tuple destructuring.

[jtobin]

Story checks out -- I observed this when using -Aon an Arvo directory that included the ;; changes, but the pill used was indeed the standard 0.7.4 one grabbed from bootstrap.urbit.org. I'll nuke this one.