Re: [Xsb-development] trie code

[Bart D. <Bar...@cs...>]

thanks for the explanation

> I think this a problem for the "no_cp" and "trust" instrs.  Shouldn't
> these use conditional trailings?  If so, the same problem is faced by
> these types of instrs using the other unify_with_trie_XXX macros.

the unify_with_trie_XXX macros use conditional trailing, but because
their condition (ebreg) is sometimes messed up immediately before
by trie_assert_inst, they can trail more than they should - it's not
their fault though, the problem is with the save_find_locx(ereg);
in trie_assert_inst (I incorrectly diagnosed this redundant trailing
as unconditional, because I am not familiar with the trie code and
because I read some comments in tst_retrv.c which are apparently
unrelated)

I am not qualified to run the test suite
if someone has the time and patience to do so, please remove the
save_find_locx(ereg); in trie_assert_inst, run the test suite and tell
us about any NEW problems; if there are no NEW problems, I suggest
this change is committed

if there are NEW problems, I suggest that we tackle these separately
because the save_find_locx(ereg); in trie_assert_inst causes a
problem, I cannot believe it is correct and it might be there just to
cover up for some other bugs; but I would like to hear confirmation
from more knowledgable people

cheers

Bart