Re: DRL updated!
"Arkady V.Belousov" <[email protected]>
| Newsgroups | gmane.os.freedos.devel,gmane.spam.detected |
|---|---|
| Message-ID | <[email protected]> |
X-Comment-To: Michal Tyc Hi! 23-Окт-2002 15:01 [email protected] (Michal Tyc) wrote to [email protected]: >> MT> 1. It is not re-entrant. Try to forget about it and find the source >> Not necessarily: make CLI before modifying INT instruction and all will >> work perfect. Should I proof this for you? MT> No, it's obvious. Of course, I can always say that preventing re-entrancy MT> and making the code fully re-entrant is somewhat different, but let's MT> don't start flame wars ;-) I don't intend to start flame wars, but I should mention, that CLI around 2-5 instructions is not equal to "preventing reentrancy" and this is reasonable compromise between "full reentrancy" and code/data compactness. >> Yes. :) But BC RTL prepares INT instruction in array on stack. This >> solves both problems (reentrancy and ROMmability). MT> Good solution, indeed. It can even work in protected mode when code MT> and stack segments overlap. If I right understand, then in this case you shold know at which CS offsets reflected SS' addresses to insert this offset into "CALL CS:[bp+-offset]"?