00:00 |
billymg |
finally got trb to build on my 2022 musl gentoo, using jfw's system compiler method with some minor patches for issues i bumped into. full writeup coming soon |
| |
~ 44 minutes ~ |
00:45 |
shinohai |
neato billymg will read |
| |
~ 3 hours 2 minutes ~ |
03:47 |
signpost |
in my own adventures on pentacle, got the vpatches separated out today. http://paste.deedbot.org/?id=feQp |
03:48 |
signpost |
I took care to separate the original src vs fixes, and provided sha256 of src origins where there was a canonical artifact released from "upstream" |
03:49 |
signpost |
for example: 720778 signpost automake_1_15_1 Genesis derived from automake-1.15.1.tar.gz (sha256: 988e32527abe052307d21c8ca000aa238b914df363a617e38f4fb89f5abf6260) |
03:52 |
signpost |
kinda mindboggling eh? 1.1gb of *text* for the kernel |
04:00 |
signpost |
curious what mass of that is drivers, as I'd like to do a full readthrough at some point. gcc as well. |
04:03 |
signpost |
looking at it all, it does make one want to get sbcl to bootstrap with tcc frankly, but I'm not eager to reinvent userland utils any time soon. |
04:04 |
signpost |
iirc there were kernel experiments with tcc, but I imagine by now that'd incur a bunch of porting w/e gcc-10000.0 extensions they've started using, or lose access to more recent drivers. |
04:05 |
signpost |
at any rate, this pig saddle's shaping up. |
04:05 |
* |
signpost afk |
| |
↖ |
| |
~ 1 hours 44 minutes ~ |
05:49 |
billymg |
test2 |
05:49 |
billymg |
ok, that's back |
06:01 |
billymg |
the one (luckily) dropped line, right after this line: <asciilifeform>: signpost: actually last asciilifeform knew, kernel used 0 gcc5isms or anyffin of the kind. would actually be surprised if didn't build w/ tcc even nao w/out heavy massage. |
06:01 |
bitbot |
Logged on 2022-01-30 04:05:17 signpost[asciilifeform]: afk |
| |
~ 10 hours 15 minutes ~ |
16:17 |
asciilifeform |
ty billymg ! |
16:18 |
signpost |
asciilifeform: interesting, that might be worth an experiment |
16:19 |
asciilifeform |
signpost: pretty sure all of my current kernels built w/ 4.9. not tried tcc tho |
16:25 |
signpost |
they did add a requirement for gcc-5.1.0 very recently, but before that 4.9 works |
16:25 |
signpost |
https://github.com/torvalds/linux/commit/76ae847497bc5207c479de5e2ac487270008b19b |
16:25 |
* |
signpost has been collecting these for an eventual pentacle poast |
16:26 |
signpost |
5.14 is recent enough that backporting drivers wouldn't be nightmare, though. |
| |
~ 2 hours 33 minutes ~ |
19:00 |
awt |
asciilifeform: in the common prologue there is perhaps a missing case: that of the first direct or broadcast message received, which will have no antecedent. These messages should not be added to the order buffer. |
19:00 |
asciilifeform |
indeed |
19:06 |
awt |
I believe this requires that we check for and reject any subsequent messages from a handle with a 0x00 self chain value that has already been seen. |
19:06 |
asciilifeform |
awt: why would this require a special case? |
19:06 |
asciilifeform |
if you've been seen, there's a handle chain for you |
19:07 |
asciilifeform |
and subsequent appearance of same handle 'as if born' are obv. bogus. |
19:17 |
awt |
So assume we have received first message from alice. Alice sends a second message with selfchain = 0x00. Antecedent check by selfchain is pointless since 0x00. Then we have to check ... long buffer? for any previous messages from Alice. If there are, reject. |
19:17 |
asciilifeform |
awt: which variant of algo are we discussing ? cuz under this one, it's trivial |
19:18 |
asciilifeform |
i.e. we know whether we've already seen a speaker 'alice', at all times |
19:19 |
awt |
this one: http://www.loper-os.org/pub/pest/pest_draft.html#431-common-prologue-for-all-packets |
19:19 |
asciilifeform |
a |
19:19 |
asciilifeform |
imho that one oughta go, and replace w/ above |
19:20 |
asciilifeform |
on acct of this and other gnarl cases |
19:20 |
* |
asciilifeform sadly not had chance to write variant of draft w/ the algo proposed above just yet, hands 100% full |
19:21 |
* |
awt will perhaps skip this case for now then |
19:22 |
* |
awt is eager to release some working version of GetData for the sake of log integrity |
19:24 |
asciilifeform |
awt: imho the Right Thing re getdata is to search the in-memory buffers 1st, followed by disk log (down to a time T set by operator) |
19:24 |
asciilifeform |
the handling of getdata's i dun expect will change much in following revs of spec |
19:24 |
awt |
yeah currently implementing LongBuffer |
19:25 |
awt |
released blatta stores it all in the log table - pretty messy |
19:25 |
asciilifeform |
if you 1) received a getdata 2) have $msg 3) have spare cpu cycles -- you send the msg to the requester. |