melges-morgen/gsoc-gcc-regex

Repository to work on the Regex in gcc

gsoc-gcc-regex's Issues

_Scanner passes incorrect expression

The scanner skips, but should not, this expressions:

• "["
• "+*"
• "a?+"

Fix test fails in common/weekly and perform comparison between common/weekly and master.

Add check of correctness of char interval expressions

It needs to fail on patterns like "[c-a]" (which isn't valid).

Scanner fails on correct expressions

Scanner fails on correct expressions like "[ca-]" which accepts "a", "c" and "-".

Merge commom/weekly into gcc4.8 trunk.

Segmentation fault in TokenFactory

If you try to compile a regular expression \[a-z\] program send SIGSEGV.

Program received signal SIGSEGV, Segmentation fault.
0x000000000040b8b5 in __gnu_cxx::operator!=<std::__regex::_TokenFactory<std::regex_traits<char> > const*, std::vector<std::__regex::_TokenFactory<std::regex_traits<char> >, std::allocator<std::__regex::_TokenFactory<std::regex_traits<char> > > > > (
    __lhs=<error reading variable: Cannot access memory at address 0x7fffff7feff8>, 
    __rhs=<error reading variable: Cannot access memory at address 0x7fffff7feff0>)
    at /home/melges/Programs/gccRegexWorkspace/current-gcc-build/include/c++/4.8.0/bits/stl_iterator.h:825
825     operator!=(const __normal_iterator<_Iterator, _Container>& __lhs,

In backtrace more then 7000 calling to _M_alternative.

From libstdc++-v3/testsuite/28_regex/algorithms/01_compilation/01_basic_valid.cc:11

The same for (\[\[a-z\]\])*, (ss)s{s,s}\[dd\],

_M_scan_in_bracket failed on valid expression

_M_scan_in_bracket failed on (([a-c])+) expression: after analysis of [a-c] token _M_curToken is _S_token_collem_single but _S_token_subexpr_end was expected.

UPD 13.07.12
_M_scan_in_bracket fails on [a-z-A-Z]

Add D_GLIBCXX_DEBUG

Add debuging stream, info http://gcc.gnu.org/onlinedocs/libstdc++/manual/debug_mode.html .

Add handling of escape symbol in _M_scan_in_bracket

_M_scan_in_bracket must be able to handle expressions like this []].

ECMA Script Support

Add ECMAScript support to _M_scan_in_bracket and _M_bracket_expression.

_M_eat_escape fails when after \\ be a regular character

_Scanner<char const*>::_M_eat_escape fails on ((ss}}\a\s near \ and fails on ++?+??+?

_Compiler fails on correct expressions

Fails on a\{,4\}, a\{2,4\}, a\{2,\}.

Add pathalogical locale tests

locales, perforamance, measuring

Should all the scanner tokens be pre-cached? Right now, many of the tokens are using std::locale:: facets to get a result, and throwing away that locale-processed result. In other areas of the library, using locales without caching is slow.

Think about this issue from a design standpoint. At what point are things cached? What difference does it make? If we cannot measure any performance problems, then it doesn't make sense to measure it.