freenode/#sbcl - IRC Chatlog
Search
11:57:49
scymtym
dougk: i think the "don't know how to dump #<TN>" problem also affects builds with SBCL 1.3.1.debian host
13:41:39
dougk
scymtym: i wonder wtf they did our build options! is it in make-host-1 or make-host-2 ?
13:42:50
dougk
pfdietz: the GC flaw? after the first GC that mistakes an offset for an address, the function it touched is completely unusable
13:45:16
scymtym
dougk: full build log: https://techfak.de/~jmoringe/sbcl-1.3.1.debian-build-host.log
13:46:55
dougk
i'd have been even more worried if aspects of the host (though it be our own code) affected make-host-2, which is exactly what we claim to insulate against
13:48:23
dougk
i recognize this problem as a bugfix i put in. the timeframe seems wrong for it to have been fixed post-1.3.1 though
13:53:58
dougk
so ... the upshot is that due to a possible "bug" (or not), in _any_ Lisp compiler that believes that (SOMEFUN +foo+) must dump the *literal* value of +foo+ where +foo+ was defined by (defconstant-eqx +foo+ (hair) #'equalp), we can not use defconstant-eqx as it was intended.
14:03:29
dougk
probably just have !possibly-defconstant-eqx never become a defconstant on the host. It's documentation mainly.
14:06:15
dougk
so the interesting thing about this is that for all the users who think they're getting some value out of alexandria::define-constant in that you can pedantically specify any comparator, do they realize that fundamentally the spec is deficient / not restrictive enough about what compilers can do, and they might end up with N copies of the constant that aren't equal under the comparator ?
14:16:47
scymtym
yeah, seems like an invitation for very subtle bugs, potentially depending on details of the compilation setup and the implementation used
14:44:34
pfdietz
Ok. I didn't have that URL, so I searched at sourceforge for sbcl. It did not come up.
14:53:09
|3b|
looks like it doesn't show up with their default filter of 'windows' (or mac if that's your default)
15:38:10
|3b|
from main page, searching for "sbcl", on results page right under the text entry box for search term , clicking "Filters * Windows" lets me change it
15:47:14
joshe
maybe someone could edit the platform list on the sourceforge project to include windows
17:29:46
scymtym
seems to a permission problem. probably related to sourceforge changing things: https://twitter.com/sourceforge/status/963684868621242368
17:33:29
scymtym
but the window for that should be basically the same as before, assuming you updated the issues before pushing
17:37:57
jackdaniel
scymtym: I think that if you merge changes hashes of your commits doesn't change (in contrary to rebase) - am I wrong? need to read more about git in that case
17:52:40
stassats
i pestered jackdaniel to fix one conformance bug, but now i don't know what else is broken