The hypothesis is that by implementing failure by returning NULL (instead of setjmp/longjmp)
we will get better (= faster) code from the C compiler.

Submitted on 5 June 2005 at 22:09

On 26 August 2005 at 09:33 Jira commented:

STR-312, visser:
it’s done and the hypothesis seems to be confirmed; new scheme is faster than old one

Log in to post comments