ff11ba9847
Updated copyright year to 2015
2015-01-03 12:14:58 +01:00
62d07878d7
Fixed two issues in matchcompiler:
...
- Use correct findmatch or findsimplematch for verification
- Print failing pattern also for findmatch/findsimplematch
2014-07-02 15:25:54 +02:00
ec59f1d9df
matchcompiler.py: Check Token::type() before comparing Token::str() against "||" or "|"
2014-06-30 14:04:18 +02:00
78f0a47170
matchcompiler.py: Show pattern that caused the InternalError
2014-06-30 14:04:18 +02:00
47a1122eb8
Updated matchcompiler to latest changes
2014-06-14 10:57:38 +02:00
a3d1cac79c
Ticket #5373 : Keep track whether tokens come from a C or C++ file to properly handle %type% in Token::Match with "delete"
2014-03-08 21:26:06 +01:00
a9df4a3288
man, htdocs, gui, test, tools: bump a few more copyrights to 2014
2014-03-03 11:03:17 +01:00
398ed973f4
matchcompiler: autopep8 formatting
2013-10-13 11:01:50 +02:00
45cf562212
matchcompiler: fixed find(simple)match compiler problems.
2013-06-02 13:57:25 +02:00
35b13da3a5
matchcompiler: find(simple)match doesn't work well. the compiled functions are not static => the wrong function can be executed. this is a quick fix that just disable find(simple)match compiled patterns.
2013-06-02 12:41:34 +02:00
650027a07a
Fixed #4797 (matchcompiler.py cannot handle overloaded findsimplematch and findmatch)
2013-05-15 18:18:58 +02:00
ae73466f97
Python scripts: PEP8 fixes
2013-03-03 09:59:16 +01:00
2ad38c8701
Updated matchcompiler to support %cop%
2013-03-01 02:52:44 -08:00
4fb6751d08
MC: Enable compiled matches for Token::findmatch() and Token::findsimplematch()
...
Speed is exactly the same and the --verify mode of the
match compiler ensured everything works as expected.
2013-01-16 21:50:32 +01:00
9f9c654621
MC: Implement verify mode for compiled Token::findmatch patterns
2013-01-14 11:18:33 +01:00
63e18d6b1c
MC: Split _replaceTokenFindMatch() into two small functions
2013-01-14 11:06:59 +01:00
017783a77e
MC: Save one indentation level
2013-01-14 11:06:55 +01:00
3bb8237cdc
MC: Turn MatchCompiler tests into python unit test
2013-01-14 11:06:33 +01:00
c0044122f4
Add self test for _replaceTokenFindMatch
2013-01-12 15:28:38 +01:00
ff9e864a07
Add self test for replaceTokenMatch
2013-01-12 15:28:38 +01:00
a7377f180a
Quote output of _assertEquals()
...
It's easier to detect trailing spaces this way.
2013-01-12 15:28:38 +01:00
4038d4b562
Place main code into own function
2013-01-12 15:28:38 +01:00
fdcfbd5ff9
Add missing copyright header
2013-01-10 11:10:45 +01:00
6236e1dc38
Token::Match: removed harmless special code for initial '!!foo' patterns.
2013-01-09 20:13:33 +01:00
de36324d62
Match compiler: Add cmdline handling
2013-01-09 19:41:03 +01:00
97e1e50139
Match compiler: Workaround broken code generation for "!!foo" patterns
...
This is just meant for the upcoming release.
2013-01-09 19:25:24 +01:00
e1be85a552
Match compiler: Be more user friendly on how to invoke it
2013-01-09 19:22:40 +01:00
ade0b741b7
Match compiler: Create 'build' directory if needed
2013-01-09 19:15:13 +01:00
982503f457
Match compiler: Workaround broken optimizations in verify mode
...
If the match compiler uses the 'verify' mode and we
compile with -O2, some tests comparing the on-the-fly-parser
to the compiled match fail.
Small functions are inlined by the -O2 compile flag.
If we disable function inlining and still compile with -O2,
everything is back to normal.
gdb didn't show anything useful during the mismatch
since the needed variables are optimized out. Once
we start printing them, the problem vanishes, too.
-> Can only be diagnosed at the x86 assembly level.
The problem vanished by switching the invocation order
of Token::Match() and the compiled match, so just swap them.
Also add commented out helper code
to better diagnose mismatch problems.
2013-01-09 19:03:42 +01:00
9970414a2f
Match compiler: Implement 'verify mode' for Token::Match
...
Verify mode runs the same ::Match pattern on a token
using the on-the-fly parser and the compiled match parser.
The result is compared and we abort on mismatch.
Already detected an error in our test suite, needs investigation.
2013-01-07 22:27:57 +01:00
9b51d81e99
Match compiler: Split _replaceTokenMatch() into two small functions
2013-01-07 21:39:49 +01:00
3918948dd5
Match compiler: Aggregate match functions with the same pattern / function call signature
...
Reduces the number of generated functions from 2290 to 1916.
2013-01-04 04:47:01 +01:00
5f09cb2e4d
Match compiler: Store matchFunctions and matchStrs in the class instead of passing it around
2013-01-04 03:56:21 +01:00
542c6e82d3
Match compiler: Cosmetic changes after running pylint
2013-01-04 03:44:04 +01:00
71a236b3df
Match compiler: Turn code into a python class
...
This will make passing around internal states a lot easier
2013-01-04 03:38:40 +01:00
f9c752667e
Match compiler: Inline generated pattern match code for Token::findmatch
...
Verified the output in build/ stayed the same
if Token::findmatch support is disabled.
2013-01-04 03:13:22 +01:00
4b386986fc
Match compiler: Check varid on first use like Token::Match does
2013-01-04 01:14:52 +01:00
5d5efdf99f
Match compiler: Add support for Token::findmatch
...
(still disabled for performance reasons)
2013-01-04 01:01:57 +01:00
7417144784
Match compiler: Compile Token::findsimplematch() calls
...
Disabled for now as it runs slowers than before.
2013-01-03 19:33:37 +01:00
62b05193c4
MC: Split convertFile() function into smaller pieces
...
Output of build/ stayed the same.
2012-12-31 15:30:27 +01:00
311630ac1c
Match compiler: Python syntax simplification
...
Output in build/ stayed the same
2012-12-31 15:03:00 +01:00
51cff0cb9b
Match compiler: Speed up Token comparison with string literals. Ticket #4451
...
"Cache" string literals as C++ std::string objects.
The string length gets cached and so we don't need to
call strlen() on the literals at runtime.
2012-12-31 14:45:16 +01:00
3c915b534d
Match compiler: Collect and cache plain C-strings
...
Profiler runs with gperftools showed we were spending
a lot of time in strlen() even with compiled match patterns.
If you compare a std::string with a plain C-string,
the comparison operator needs to calculate the length
of the C-string via strlen().
We can avoid this by turning all C-strings into C++ strings,
which cache the string length internally.
The match compiler has been adapted to collect all
C-strings during compilation and aggregate them.
Benchmark on the 'rpm' code base on a Core i7 920 box:
cppcheck without matchcompiler:
real 0m30.977s
cppcheck with previous matchcompiler:
real 0m28.157s
cppcheck with cached C-strings:
real 0m17.823s
2012-12-30 23:24:28 +01:00
fe5ee76165
Match compiler: No need to handle %num% twice
2012-12-30 20:33:49 +01:00
3d0b4a9f7e
Add execute permissions to matchcompiler.py on linux
2012-12-25 13:18:24 +01:00
a2885a1759
Python 3 compatibility for the helper scripts
...
print() is a function in python 3.
Works with pyhton 2.x, too.
2012-12-25 13:12:58 +01:00
3b3fe1c616
matchcompiler: throw InternalError if Token::Match is called with the varid 0
2012-12-19 06:45:05 +01:00
b6786c87b4
matchcompiler: handle %varid%
2012-12-14 16:59:37 +01:00
2d8ed65d43
compiled Token::Match patterns: refactorings,improvements
2012-12-10 20:28:11 +01:00
a11ec4a5ee
matchcompiler: replaced regular expression with simple code that parses the code better
2012-12-08 19:29:16 +01:00
Daniel Marjamäki
PKEuS
PKEuS
Simon Martin
Matthias Krüger
Frank Zingsheim
XhmikosR
Thomas Jarosch
Edoardo Prezioso