Search
Tuesday, 4th of December 2018, 17:30:09 UTC
3:58:09
beach
Good morning everyone!
3:58:17
no-defun-allowed
morning beach
4:47:32
beach
So, the situation is not so good at the moment.
4:48:00
beach
During bootstrapping, a large number of funcallable standard objects are created.
4:48:36
beach
Apparently, SBCL does not have enough space available for those objects.
4:48:58
beach
There is no startup switch to change the size.
4:49:47
beach
And I tried the compile-time options suggested by Douglas Katzman in this message from 2 years ago: https://sourceforge.net/p/sbcl/mailman/message/35497099/
4:51:22
beach
I also tried changing to CCL. I get a mysterious error early on in the bootstrapping procedure that a slot reader is called with 1 argument but requires at least 2. I think I would need some help from rme to figure that one out.
4:52:02
shka_
beach: 'enough space'?
4:52:33
beach
Apparently, funcallable standard objects are allocated in "immobile" space.
4:52:49
shka_
therefore --dynamic-space-size can't help?
4:52:51
beach
And the amount of immobile space is not enough.
4:53:05
beach
Correct, I already have a lot of that.
4:53:25
shka_
ACTION is pretty sure that he has more ;-)
4:53:52
shka_
maybe there is just one magic number in the sbcl source that controls that
4:54:04
beach
If so, I think stassats would have told me.
4:54:13
shka_
then it would simply boil down to recompiling sbcl
4:54:48
beach
Instead, he changed something in the SBCL code, asked me to download a fresh version, compile it, and install it.
4:54:54
beach
But the same problem persists.
4:55:36
shka_
well, i will check anyway
4:56:34
shka_
i will grep sbcl source code for something like +immobile-space-size+
4:56:53
beach
I already did that, but I could have missed something.
4:57:03
beach
I am fairly sure that stassats would not miss it though.
4:57:04
shka_
i don't expect to find anything, but you never know
4:58:08
shka_
beach: one thing: you surely downloaded master from github?
4:58:40
shka_
ok, i bet that this is the commit https://github.com/sbcl/sbcl/commit/e7cd2bd40f5b998821874b1720d61735f2cf292e
4:59:44
beach
Let me check that this code exists in what I downloaded.
5:01:20
beach
It was supposed to have been removed.
5:02:49
beach
I'll try removing that code and see...
5:03:13
shka_
there is TODO for specyfing inmmobile space size on startup
5:04:30
beach
I cloned SBCL for the first time after stassats told me to do it.
5:04:39
beach
I don't see how a pull could do any better.
5:04:50
beach
Oh, that would be nice.
5:06:01
shka_
i think that it is possible to specify immobile page size in immobile-space.c
5:06:31
shka_
this makes me to think so #define WORDS_PER_PAGE ((int)IMMOBILE_CARD_BYTES/N_WORD_BYTES)
5:06:53
shka_
file also contains aforementioned TODO note
5:07:47
beach
Thanks. I'll try stassats' fix first. Then this one.
5:08:22
shka_
well, stassats fix is probably a fix while this would be ugly hackery ;-)
5:09:08
shka_
it is funny to see references to PCL in the source code
5:09:15
shka_
sbcl goes waaaaaaay back
5:10:27
shka_
how foolish to doubt him ;-)
5:11:33
beach
Thank you for finding this for me.
5:11:42
beach
I guess he must have committed to a different branch.
5:16:06
shka_
beach: could you check if this commit even exists in your clone?
5:16:19
shka_
trying to checkout to it would be sufficient
5:16:28
beach
I am pretty sure it does not.
5:16:42
beach
I don't know how to do that.
5:17:36
shka_
well, don't bother in that case but git checkout e7cd2bd40f5b998821874b1720d61735f2cf292e
5:18:29
beach
fatal: reference is not a tree: e7cd2bd40f5b998821874b1720d61735f2cf292e
5:19:30
beach
I need a break. Back in 30 minutes or so.
5:19:44
shka_
maybe this has to do something with github mirror or something that delayed code distribution just enough for you to pull older version
5:20:09
shka_
does not matter anyway
Wednesday, 5th of December 2018, 5:30:09 UTC