| Commit message (Collapse) | Author | Age | Files | Lines |
... | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | | |
Remove the Bison-bdidge kludge.
Rearrange yyguts_t so the public members are grouped together.
|
| | | |
| | | |
| | | |
| | | | |
...to clean out a forward declaeation.
|
| | | | |
|
| | | |
| | | |
| | | |
| | | | |
Not touching the public interface yet.
|
| | | | |
|
| | | |
| | | |
| | | |
| | | |
| | | | |
It was fixed up by generating matcing unbalanced markuo in dfa.c,
but a good idea this was not.
|
| | | | |
|
| | | |
| | | |
| | | |
| | | |
| | | | |
No publicly visible names are touched yet. This is all preep work.
All tesrs pass.
|
| | | |
| | | |
| | | |
| | | | |
...presebtly commented out. All tests pass.
|
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | | |
Begin the Go back end. Right now it's a fake that generates C...
...but this commit puts all the build and testnmachinery in place
for when it's not. All tests pass.
|
| | | | |
|
| | | | |
|
| | | |
| | | |
| | | |
| | | | |
Due to some limitayions in git ignore wildcarding.
|
| | | | |
|
| |\ \ \
| |/ / /
|/| | |
| | | | |
All tests pass.
|
| | | | |
|
| | | | |
|
| | | |
| | | |
| | | |
| | | | |
...and fix the bug that turned up.
|
| | | | |
|
| | | |
| | | |
| | | |
| | | |
| | | | |
This reduces the complexitry of the test machinery as well as making
tests less backend-specific.
|
| | | | |
|
| | | |
| | | |
| | | |
| | | | |
Also, yyinpt.txt is generated and does not need to be checked in.
|
| | | |
| | | |
| | | |
| | | |
| | | | |
It decayed because the test for this feature only exercised the nr
back end. We add a ruleset--based test to verify on all back ends.
|
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | | |
We a ruleset test for BOL matching and a test for the array case of yymore.
Improve yymore testing by adding tests that require a BOL match in
both array and pointer cases.
Also, convert lineno sporadic test to be ruleset-based so it's
applied on all back ends.
No cide changes.
|
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | | |
Remove two dependencies that the C99 back end no longer has.
Repair to force in-time building of skel.h files. This way of doing
it reduces where a new skel dependency has to be declared in the
vuilld recipe to the one spot where the flex.h file is made from the
flex.skl file.
The posixly correct test only needs to be done once, since it only
affects how the tables are genetated rather than setting code
conditionals in the back end.
Also, some comment typo fixes.
|
| | | |
| | | |
| | | |
| | | |
| | | | |
..,without resorting to forward declarations. Last C99 cleanup before
the Go port begins.
|
| | | | |
|
| | | | |
|
| | | |
| | | |
| | | |
| | | | |
...for automatic remake when a new one is added.
|
| | | | |
|
| | | | |
|
| | | | |
|
| | | | |
|
| | | | |
|
| | | |
| | | |
| | | |
| | | | |
One less C-ism to deal wit h when porting to a new target languages.
|
| | | |
| | | |
| | | |
| | | |
| | | | |
Also, roperly condition some variables related to trailing contexts.
Makes generated parsers ever so slightly smaller in this case.
|
| | | | |
|
| | | | |
|
| | | |
| | | |
| | | |
| | | |
| | | |
| | | | |
There's only one left. in the prefix code. We leave that in place
because ity will be needed if we ever want to replace the default
C back end with C99.
|
| | | | |
|
| | | |
| | | |
| | | |
| | | |
| | | | |
The functions are: yy_current_buffer(), yy_flush_current_buffer(), and
yy_set_interactive().
|
| | | | |
|
| | | | |
|
| | | | |
|
| | | | |
|
| | | | |
|
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | | |
Ensure that derived skel.h includes are removed on every clean.
Fix a bad shell line that prevented proper cleaming of derived *.c
files from tests.
|
| | | |
| | | |
| | | |
| | | |
| | | |
| | | | |
There seems to be something in the build recipe or test system that
makes the yyreject test unreliable. Reverting until I figure out
what.
|
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | | |
The tradeoff is a bit delicate here; yy_set_bol() did not strictly
speaking need to change. On the other hand, squeezing the underscores
out of YY_AT_BOL() but not yy_set_bol() would have been exactly the
kind of irregularity that people trip over and curse.
The underlying issue is of course languages like Go and Python with
validators that get upset by all caps and embedded underscores. On the
one hand, we want to not be churning the Flex API gratuitously. On
the other hand it is desirable if as few entry point names as possible
need to change in different target languages.
The approach I'm taking to deciding this question in individual cases
is this: if an entry point name is likely to appear in actions, and
must be recognized as magic by Flex itself in order to have a final
context argument added or an object reference prepended, then it must
be multilanguage-safe - not all-caps and no underscores.
Any name that does *not* need to be wired into Flex can be left
alone. Users may have to change it when they port.
|
| | | | |
|