| Age | Commit message (Collapse) | Author | Files | Lines |
|---|
|
Fixes building tarballs.
We also include the test suite, the original stand-alone Makefile and
READMEs.
|
|
|
|
* This was a combination of conversion loss and a too small color map.
* The test suite was actually broken for Unicode since we have to
`setlocale(LC_ALL, "")` for the libc Unicode handling functions
to work as expected. SciTECO already did that.
* Added test cases for plain Unicode (Cyrillic) characters in the
regexp and for character ranges.
|
|
This allows you to pass in additional flags like `-DREG_DEBUG`.
terex is *not* built with assertions by default unless you define REG_DEBUG.
|
|
|
|
Allows you to set up a callback that's invoked repeatedly during
matching. If it returns non-0 matching will abort and REG_EINTR is
returned.
Allows you to interrupt long-running regexp executions that would
otherwise block the UI or hang the entire application.
|
|
* Perhaps I am also obliged to document what I changed in the copyright header?
Not sure.
* The license header doesn't necessarily have to be reproduced in binary
distributions.
|
|
tere_error() for consistency
The headers still define regfree() and regerror() functions,
but they shouldn't collide with libc unless you include the POSIX regex.h.
Also tere_free() can now be called on nullified regex_t objects, which
eases error handling a bit.
|
|
I put them into longest() as well, but it appears that both find() and cfind()
will always try shortest() first and should abort if that returns NULL.
|
|
* So you can anchor executions of already compiled patterns.
In principle REG_BOSONLY could also be handled by the same code.
* Required by SciTECO.
* Test suite has been fixed and extended.
The test program now takes getopt() style short arguments to specify flags.
|
|
Fixes inclusion into programs that also include unistd.h on BSD,
which also provides re_comp() and re_exec() implementations.
|
|
|
|
I missed the possibility to query the compilation flags from the
execution variables.
|
|
The original hsrex README has been re-added as README.old.
|
|
compile-time flag
You no longer need to build two libraries just for supporting raw/ASCII and UTF-8
patterns.
This makes using the library a lot easier and reduces the total binary size.
Since strings are always `unsigned char *` now internally and the raw vs. UTF-8
decision is important only in a few select places, it would make no sense to
use meta-programming techniques.
The test suite has been extended and is now fixed for cases with embedded
non-printable characters.
|
|
Builds two convenience libraries: libterex.la and libteurex.la.
|
|
This will be the SciTECO regular expression engine and we have to fork
it permanently.
|
|
|
|
* They are built with `-DREGEX_UTF8` instead of `-DREGEX_WCHAR`.
Functions are called reg_ucomp() and reg_uexec() instead for consistency.
The library is now called libhsurex.so instead of libhswrex.so.
* The `chr` type is now always `unsigned char`.
As a result many other uses of the `chr` type had to be changed to pchr
(which is always large enough to hold a byte or wide character).
Generally we try to keep code changes as small as possible since
we may have to backport changes from the Tcl codebase or contribute
patches to the Tcl project.
|
|
|
|
|
