Benito van der Zander
Benito van der Zander
When I disable the assembler code there, it blows up in SearchMatchFullReversed: ``` F/libc ( 1556): Fatal signal 11 (SIGSEGV) at 0x00000400 (code=1), thread 1556 (ibela.videlibri) I/DEBUG ( 792): ***...
fpc 3.1.1 (I think I compiled it on Apr 5 2015)
It still happens with new fpc (and disabling the assembly blocks fixes ti) ``` 05-09 15:55:54.897 4119 4119 F libc : Fatal signal 11 (SIGSEGV), code 1 (SEGV_MAPERR), fault addr...
Hey, look a follow up: https://arxiv.org/abs/1611.00918 With an even faster algorithm for `(foo|bar|123|etc)*`
Now I understand the problem. If `.` could match anything and you have `>..ä
And since the expansion is simpler, it is probably faster, too
> But I think, I'll add a extra "all-byte-match" token, which matches always then every byte/codeunit, in the UTF8 and also in the non-UTF8 mode. Or special byte level character...
https://bugreports.qt.io/browse/QTBUG-46892
>I also feel that we're trying to impose a symmetry between maps and arrays where possibly none exists. In a theoretical sense, an array is just a map with integer...
These Unicode tables are also very big Can't have a lean, self-contained implementation if it has to carry these tables around