From: <wil...@ai...> - 2007-07-01 18:19:45
|
On Thu, Jun 28, 2007 at 07:11:15AM +0100, Christophe Rhodes wrote: > wil...@ai... (William Harold Newman) writes: > > > However, safe cheap trivial changes though they are, they > > also seem to come with a global requirement that SBCL's code continue > > to conform to the ANSI-Standard-as-revised-by-Allegro in order to > > avoid code rot. Enforcing that seems to involve manual tedium or extra > > automated checks at compile time. > > I'm inclined to agree that "modern"ising the code isn't a great > motivation for frobbing the codebase, but there is a possible other > motivation for at least some of the changes that Marco has sent in: at > the moment, SBCL fails to bootstrap if the builder has altered > (readtable-case *readtable*) in their image. True, but is there something special about tweaking READTABLE-CASE compared to setting *READ-BASE* to 16? It's not just a rhetorical question --- for all I know, there is some formal fundamental divide where READTABLE-CASE is on one side and *READ-BASE* is on the other. But my informal impression was that software can reasonably expect both of them to be in the default setting. > This doesn't involve changing (intern "FOO" "BAR") to (intern '#:foo > '#:bar), where I have an irrational hatred for the second form; it > does involve changing various VOPs where currently we have > > (inst foo) > LABEL > (inst bar) > (inst jmp label) In this case, writing this as > (inst foo) > label > (inst bar) > (inst jmp label) seems more natural to me; the old way isn't unnatural enough that I was motivated to change it, but I'd consider the all-one-case form to be a (very tiny) improvement. -- William Harold Newman <wil...@ai...> PGP key fingerprint 85 CE 1C BA 79 8D 51 8C B9 25 FB EE E0 C3 E5 7C "If you could have one wish, what would it be?" "I wish that all Web browsers would drop their proprietary tags and conform to the W3C's HTML 4.01 and the ECMA Script 262 v.2 specifications." -- <http://hackles.org/cgi-bin/archives.pl?request=17> |