| Commit message (Collapse) | Author | Age | Files | Lines |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
After processing the last item for a named machine, mark only those items with
the same ParseData pointer as processed. Previously, we would mark all
preceding items as processed. This is wrong and is a problem for sequences
like:
machine A // some defs
machine B // some defs
machine C // some defs
machine A // last
machine B // last
machine C // last
Once the second A block is seen, then the first A, B and C would all get marked
as processed and ready for writing, even though B and C were not complete yet.
Also, now processing items with no ParseData pointer earlier by immediately
marking them as processed and moving the last flushed pointer forward even if
we didn't process an item with a parse data pointer.
|
|
|
|
|
|
| |
If the .la files for libcolm and libfsm are not present then fallback to using
AC_CHECK_LIB for these libraries. Many packaging systems strip out .la files so
we need to still be able to link if they are not there.
|
|
|
|
|
|
| |
With the fix in for #80, test cases that embed from-state actions, but no eof
actions, now need to have eof set. Updating output to reflect the calls on eof
is also required in some cases.
|
|
|
|
|
| |
The ts var is currently getting set on EOF due to ragel 7 executing from-state
actions on EOF. This is a regression from ragel 6. refs #76.
|
| |
|
|
|
|
|
| |
refs #80 from-state actions on EOF are executed only if there are EOF
transition actions
|
|
|
|
| |
refs #79
|
|
|
|
|
|
|
|
|
| |
* full.Dockerfile installs all dependencies, builds master and does full
testing.
* master.Dockerfile does a minimal build from master.
* release.Dockerfile builds and installs release tarballs.
|
|
|
|
|
|
|
| |
The existing dockerfile builds a release version of ragel and does not do any
testing. Renamed this to release.Dockerfile. Added a new dockerfile called
master.Dockerfile that installs all depenendencies for testing all host
languages. It then builds colm and ragel from master.
|
| |
|
| |
|
|
|
|
| |
refs #79
|
|
|
|
| |
Improved lexing of char lits and identifiers (including lifetimes).
|
|
|
|
|
|
| |
There are common regular language definitions in ragel.lm that are used in the
host-specific lexers. These need to move out into the host code so we can begin
to specialize the lexers for the different host languages.
|
|
|
|
|
| |
The language-specific parsing code should depend on the common grammar and
reduction rules.
|
|
|
|
| |
See adrian-thurston/colm#134.
|
|
|
|
|
|
| |
In most build scenarios I now find myself in, the depenencies required for
building the manual are not available. There are many uncommon packages to
grab. It seems better to build the manual only when asked for explicitly.
|
| |
|
| |
|
| |
|
| |
|
| |
|
|
|
|
|
| |
This code was present in ragel, but the colm version was used. Deleted the dead
code here an moved the real impl into InputData.
|
|\
| |
| | |
Removed -j from `make` in Dockerfile
|
| |
| |
| | |
Build process didn't seem to like the `-j` with no argument.
|
| | |
|
| |
| |
| |
| |
| |
| | |
Previously had char fixed to signed char, this is not useful on ARM as it does
not align with the host type. Instead, decide at runtime (or probably compile
time) if char is signed or not.
|
| |
| |
| |
| |
| | |
Now need to convert the buffer to a string before passing to functions that
expect string.
|
| |
| |
| |
| |
| | |
The -fPIE option is now the default and it needs to be turned off when making
simple progs that use .rodata and .text.
|
| | |
|
|/ |
|
| |
|
| |
|
|
|
|
| |
Moved make of src/include/ragel to configure script. Expect colm 0.14.7.
|
|
|
|
| |
This version of colm includes a critical fix for big-endian system. Fixes #61.
|
|
|
|
|
| |
Latest colm includes bugfixes for refcounting, which fixes a ragel issue with
includes #58.
|
|
|
|
| |
also expect colm 0.14.4, the latest release
|
| |
|
|
|
|
| |
Switching to major.minor.point format and will build some momentum around that.
|
| |
|
|
|
|
| |
Some codegen files have moved to colm, where they will be reused.
|
|
|
|
| |
Need to use unsigned min/max fields for unsigned types. Fixed size.
|
|
|
|
|
|
| |
* Checkout, build and install colm first, use when configuring ragel.
* Include test dir in SUBDIRS of main Makefile.
* Replaced colm-suite-* paths with ragel-*.
|
|\
| |
| | |
common: Fix ambiguous CHAR_MIN/MAX definition
|
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| | |
According to C/C++ standards, char is implementation-defined whether
it could hold negative values.
See: http://www.cplusplus.com/reference/climits/
In ragel char is treated as a signed value with range as [-128, 127].
This means that the CHAR_MIN/CHAR_MAX should be replaced with a more
accurate definition to align across different systems and library implementations.
Change-Id: I10668f2d2550b603101dc68f4cc1121035022abd
Signed-off-by: Jun He <jun.he@linaro.org>
|
|/ |
|
| |
|
| |
|
|
|
|
|
| |
Moved param check code to mainline as part of trimming down common file.
NameInst structs live in libfsm and can be removed from ragel.
|
| |
|