freenode/#sicl - IRC Chatlog
Search
4:44:22
beach
Besides mediocre software, the other thing I don't like is mediocre documentation. When I do `git status -s' in one sub-directory of SICL I get a T in the second column, but `man git-status' does not list T as one of the possible outputs.
4:46:23
no-defun-allowed
https://stackoverflow.com/questions/6879501/filter-git-diff-by-type-of-change#6879568 has a snippet from `man git-diff` of all places describing the letters.
4:55:01
beach
OK, I think I see what the problem is, but it required #sicl, no-defun-allowed, stackoverflow, and a second computer with a copy of the SICL repository on it to figure it out.
4:55:27
beach
Somehow, files that are symbolic links in the repository have turned into regular files.
4:56:25
no-defun-allowed
I've never checked, but the GitHub viewer shows broken symbolic links as files, just containing the pathname.
4:57:40
no-defun-allowed
I think that's how the representation could work though. If you look at the contents of those files, are they what you expect?
4:58:22
no-defun-allowed
(Not that this is what you expected at all, but do they have the contents of the linked file?)
5:00:54
beach
Yes, they are identical. Just removing the files and doing git checkout fixes the problem.
5:14:27
beach
Clearly, the software I write has defects, and the documentation is sometimes incorrect or out of date. But I am a single person, and I would think that software like GIT or Firefox would be better, given the resources at their disposal. Having said that, I think that the software that has been `released' and the documentation we produce for it, like Cluffer, Eclector, Flexichain, etc. are not too bad at all, given the resources
5:17:21
beach
I do not consider Cleavir as having been `released', since it is still being worked on. It is used by Clasp because drmeister asked to use it.
5:43:29
beach
The thread is part of the dynamic environment, so a separate register is not strictly required.
5:44:53
beach
Multiple values beyond what registers can hold can be contained there as has been suggested.
5:48:34
beach
Code that uses the floating-point register for floating-point arithmetic can just save that register temporarily.
5:49:21
beach
But since I expect most code to use integer arithmetic, or no arithmetic at all, there is a chance that this register will not have to be saved and restored very often.
5:53:18
beach
The inconvenience, of course, is that it has to be copied to a general-purpose register in order to be dereferenced.
5:56:26
beach
On the other hand, if I dedicate a general-purpose register for it, then it would very likely have to be spilled to make room for other data during the execution of a function.
6:05:08
beach
In the past I was thinking that the thread object would be accessed through the dynamic environment, like a special variable.
6:07:13
beach
And object allocation is needed for lots of stuff, including creating closures and "cells" for shared variables.
6:33:48
Shinmera
beach: Could use SSE registers instead, those would be even less likely to be used.
6:34:42
Shinmera
You might want to talk to the SBCL people though, they might have real-world based opinions on which register to pick.