freenode/#sbcl - IRC Chatlog
Search
9:40:41
flip214
mailing patches is no good, Autor: Philipp Marek via Sbcl-devel <sbcl-devel@lists.sourceforge.net> 2019-10-17 17:07:12
9:42:00
flip214
jackdaniel: well, if I put in some effort at least my name and email should be stored in git forever, right?
9:43:49
flip214
minion: memo for stassats: is it easier to merge sourceforge branches than from github? I'd like to keep my name and email in my commits...
9:57:27
minion
stassats, memo from flip214: is it easier to merge sourceforge branches than from github? I'd like to keep my name and email in my commits...
10:36:14
stassats
but don't know why your mails look like that, i applied other patches an they had proper from:
10:37:11
flip214
stassats: not officially... https://github.com/sbcl/sbcl/network/members shows quite a few users
10:38:16
flip214
stassats: yeah, perhaps SF changed the ML configuration to rewrite the "From" line ... I saw a few mails from doug with the same "via..."
10:40:21
flip214
stassats: I respect your opinion. but if you reject gitlab, github, and SF, what else is there?
10:41:25
flip214
OTOH, do you really interact with the web gui so often? Isn't the most often thing "git pull"/"git push" etc.? For that it wouldn't matter _where_ the "main" repo is
10:45:40
stassats
(SB-THREAD:CONDITION-WAIT #<SB-THREAD:WAITQUEUE {998A15C1}> #<SB-THREAD:MUTEX (free) {998A15A9}> :TIMEOUT NIL),
10:45:40
stassats
(SB-THREAD:CONDITION-WAIT 0 -438224805 :TIMEOUT #<unknown immediate object, lowtag=#b110, widetag=#x36 {98300736}>)
10:47:56
flip214
both gitlab and github _do_ link to PRs and issues... perhaps with different syntaxes (didn't check), and of course the numbering will be off for all old entries
10:49:47
flip214
still, if I had any issues with SBCL in my work life, you'd be the first I'd offer money to fix it ;)
10:52:44
stassats
maybe 0.1 is too small a timeout for the CI and CONDITION-WAIT is interrupted somewhere bad
12:39:29
stassats
these are safepoints, could it be, the interrupt hits, stores the context, then waits until safepoint, getting PC/other regs mixed up
13:25:38
pfdietz
stassats: I was assuming string= would have a special case for no keyword parameters.
13:26:29
pfdietz
It might be interesting to figure out a way to profile or scan code to rank optimization opportunities.
13:27:47
stassats
i also don't like keywordless functions, as that could be done with multiple entry points
14:06:57
pfdietz
I am sstill wondering how much chaos would ensue if the read table base preference were set to :BOTH instead of :SYMBOLS.
14:48:26
pfdietz
String constants are read as simple character strings, even if they could be simple base strings. Comparison against them should coerce them automatically at compile time to base strings if that would help.
14:48:55
pfdietz
(equal (the simple-base-string s) "foo") ==> should treat "foo" as a simple base string
14:49:43
pfdietz
Overall, there is potentially very large impact from using base strings when possible. Factor of 4 saving in space for large strings.
14:50:42
pfdietz
I wonder if it would make sense to optimize general strings for the case of strings that are mostly base characters.