ilyakooo0/urbit
1
0
Fork 0
mirror of https://github.com/ilyakooo0/urbit.git synced 2024-12-22 22:31:30 +03:00
Code Issues Packages Projects Releases Wiki Activity

clay: scry when doing file requests outside %sing

Browse Source 
In +work, we simply want to fetch file data, regardless of what the
overarching request is. (In fact, %sing would likely never hit this path
anyway.)

In order to be able to make the scry request, we track the timestamp of
the commit that contains a blob we're missing, and scry for that %da
revision.

There's two edge cases where we cannot immediately know the timestamp
that are currently assumed broken. Fix soon.
This commit is contained in:
fang 2022-02-17 15:44:16 +01:00
parent dd4a115e51
commit 3473033492
No known key found for this signature in database
GPG Key ID: EB035760C1BBA972
1 changed files with 25 additions and 14 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

39
pkg/arvo/sys/vane/clay.hoon
View File

@ -236,7 +236,7 @@
=rave
scry=(unit @da) :: timeout timer if scry
have=(map lobe blob)
need=(list [=path =lobe])
need=(list [when=@da =path =lobe])
nako=(qeu (unit nako))
busy=_|
==
@ -3159,21 +3159,27 @@
::
++ missing-blobs
|= =nako
^- (list [path lobe])
=| miss=(set lobe)
^- (list [@da path lobe])
::REVIEW should this get sorted so that we prefer getting files
:: from the latest commit, if there are duplicates?
=/ yakis ~(tap in lar.nako)
|- ^- (list [path lobe])
|- ^- (list [@da path lobe])
=* yaki-loop $
?~ yakis
~
=/ lobes=(list [=path =lobe]) ~(tap by q.i.yakis)
|- ^- (list [path lobe])
|- ^- (list [@da path lobe])
=* blob-loop $
?~ lobes
yaki-loop(yakis t.yakis)
?: (~(has by lat.ran) lobe.i.lobes)
=* lobe lobe.i.lobes
?: ?| (~(has by lat.ran) lobe)
(~(has in miss) lobe)
==
blob-loop(lobes t.lobes)
:- i.lobes
blob-loop(lobes t.lobes)
:- [t.i.yakis i.lobes]
blob-loop(lobes t.lobes, miss (~(put in miss) lobe))
::
:: Receive backfill response
::
@ -3187,9 +3193,12 @@
!(~(has by have.sat) q.q.blob)
==
::NOTE we have already fallen back to the ames case,
:: (scry responses always come in as %direct blobs)
:: so we do not need the real path here anymore.
::REVIEW right? or is this considered too fragile?
[[/ q.q.blob] need.sat]
:: actually yeah, we might retry scry later...
::TODO also wouldn't this want to check if lobe was already in need?
[[*@da / q.q.blob] need.sat]
:: We can't put a blob in lat.ran if its parent isn't already
:: there. Unions are in reverse order so we don't overwrite
:: existing blobs.
@ -3243,13 +3252,13 @@
:: :ship's remote scry isn't known to be broken,
:: or we learned it was broken more than an hour ago,
::
?: ?& ?=(%sing -.rave.sat)
?| !(~(has by sad) her)
(lth now (add ~h1 (~(got by sad) her)))
== ==
?: ?| !(~(has by sad) her)
(lth now (add scry-retry-time (~(got by sad) her)))
==
:: make the request over remote scry
::
=/ =mood [%x case.mood.rave.sat path.i.need.sat]
=/ =mood
[%x da+when path]:i.need.sat
[`- +]:(send-over-scry %back-index hen her inx syd mood)
:: otherwise, request over ames
::
@ -4846,7 +4855,9 @@
^- update-state
::NOTE putting / paths is fine, the path is only used
:: when dealing with scry responses
[duct rave ~ have (turn need (lead /)) nako busy]
::TODO but that isn't actually true, is it? we might fall back to
:: scry in the middle, right?
[duct rave ~ have (turn need (corl (lead *@da) (lead /))) nako busy]
--
::
++ scry :: inspect