11 Jan
2005
11 Jan
'05
2:19 p.m.
From: Royce Mitchell III
The only reason I didn't do something like this was because I was
worried about the effect on build-time, which at the moment
is already horrendous. If this sorting can be done quickly
( i.e. have minimal effect on build time), then I think it's
a great idea.
My gut feeling is that it won't impact build time very much, but I won't
know for sure until I've coded the thing.
Also, how much larger do you anticipate the sym files
getting, and do we have the kmode address space available
to devote to this increase in size with room to spare?
A STAB_ENTRY as currently used is 12 bytes, my proposed SYM_ENTRY is 16
bytes. There can be multiple STAB_ENTRYs for a single address, while there
will be only 1 SYM_ENTRY. Strings are handled the same by both formats, they
are stored in a string table. All in all, I'd guess the new sym files to be
10% larger than the old ones. Seems a reasonable price to pay.
I apologize if some of these questions are stupid ;)
No such thing as a stupid question. Perhaps asking on IRC whether accept()
works is annoying, but it's still not stupid ;)
Gé van Geldorp.