[Fontconfig] fontconfig: Branch 'master' - 2 commits
Raimund Steger
rs at mytum.de
Thu Mar 5 14:29:03 PST 2015
On 03/05/15 09:56, Akira TAGOH wrote:
> just committed the change. that should works better now.
Yup it's fast again now... but:
> FcBool
> FcBlanksIsMember (FcBlanks *b, FcChar32 ucs4)
> {
> - int i;
> + int lower = 0, higher = b->nblank, middle;
>
> - for (i = 0; i < b->nblank; i++)
> - if (b->blanks[i] == ucs4)
> + if (b->nblank == 0 ||
> + b->blanks[0] > ucs4 ||
> + b->blanks[b->nblank - 1] < ucs4)
> + return FcFalse;
> + while (1)
> + {
> + middle = (lower + higher) / 2;
> + if (b->blanks[middle] == ucs4)
> return FcTrue;
> + if (middle == lower ||
> + middle == higher)
wouldn't this need to be:
if (lower == higher)
Rationale:
If lower and higher are 1 apart, one bound will be checked, but not the
other. For the current list of blanks and my font cache, I observed the
following with ucs4==0xfe07 (which is /in/ the list):
sun2:fontconfig)dbx `which fc-cache`
[...]
(dbx) runargs -f
(dbx) stop at fcblanks.c:93 -if higher-lower==1 && ucs4==0xfe07
(2) stop at "fcblanks.c":93 -if higher-lower == 1 && ucs4 == 0xfe07
(dbx) run
Running: fc-cache -f
(process id 22766)
t at 1 (l at 1) stopped in FcBlanksIsMember at line 93 in file "fcblanks.c"
93 middle = (lower + higher) / 2;
(dbx) next
t at 1 (l at 1) stopped in FcBlanksIsMember at line 94 in file "fcblanks.c"
94 if (b->blanks[middle] == ucs4)
(dbx) print lower,middle,higher,b->blanks[middle],b->blanks[higher],ucs4
lower = 62
middle = 62
higher = 63
b->blanks[middle] = 65030U
b->blanks[higher] = 65031U
ucs4 = 65031U
(dbx) next
t at 1 (l at 1) stopped in FcBlanksIsMember at line 97 in file "fcblanks.c"
97 middle == higher)
(dbx) next
t at 1 (l at 1) stopped in FcBlanksIsMember at line 98 in file "fcblanks.c"
98 break;
(dbx) next
t at 1 (l at 1) stopped in FcBlanksIsMember at line 105 in file "fcblanks.c"
105 return FcFalse;
--
Worringer Str 31 Duesseldorf 40211 DE home: <rs at mytum.de>
+49-179-2981632 icq 16845346 work: <rs at interface-ag.de>
More information about the Fontconfig
mailing list