Comparing changes
Open a pull request
base repository: m-labs/flickernoise
base: 2aa0d1f
head repository: m-labs/flickernoise
compare: 152aaf7
- 17 commits
- 26 files changed
- 1 contributor
Commits on Dec 12, 2011
-
compiler: don't feed TOK_ERROR to the parser
TOK_ERROR returned after a scanner failure was still sent to the parser, forcing a parse error as well. While this didn't have any ill effects, it's better to disentangle the two.
-
compiler: report the line number and the context of an error
We'll need this when the compiler parses the entire patch, not just an expression.
-
compiler: remove trailing whitespace
Some lines had trailing tabs and spaces. Get rid of them.
-
compiler: return error string from fpvm_parse
This way, the caller has access to the full diagnostics.
-
compiler: fix multiple /* ... */ comments
The regexp was too greedy and treated /* ... */ ... /* ... */ as a single long comment. Luckily, others have already solved that braintwister for us. This is the solution by Stephen Ostermiller.
-
compiler: bring out errors encountered by things called from parser
This seems awkward but works. Maybe there's a better way. lemon documentation is silent about handling this kind of errors, and the results found when googing for lemon and YYABORT aren't encouraging.
-
compiler: moved prologue and epilogue to almost regular .fnp files
They only differ in not having the per_frame= and per_vertex= prefixes. The files are then converted to a header containing #defines of strings with the file content inside. The compiler simply calls fpvm_chunk to compile the prologue and epilogue.
-
The original MilkDrop presets are based on INF files and have a "[preset]" at the beginning that should be ignored.
-
compiler: infrastructure for supporting fragment selection prefixes
There are two changes: - instead of calling fpvm_do_assign directly, use a callback - support for pre_frame, per_vertex, and per_pixel prefixes
-
compiler: support parsing imagefileN as well
This was tricky: since file names are bare and can contain all kinds of special characters and keywords, we need to switch the scanner into a different mode, which is a little awkward with re2c.
-
compiler: dirty hacks to work around syntax problems in patches
This commit makes the parser accept the following constructs: - per_frameinit_1=... (becomes per_frame=) - per_frame= (or per_vertex; is ignored) - per_frame=//... (or per_vertex; is ignored)
-
compiler: make regression tester support per_frame and per_vertex as …
…well Trival. Should have done this earlier.
-
-
compiler: don't regularly use " in error messages; nicer format
MTK doesn't like strings containing double quotes. We can work around this by minimizing the use of double quotes sent to MTK. Since this changes all the error messages in the regression tests anyway, we can use the opportunity and also make the message format a little nicer.
-
-
We don't need it anymore.
-
This comparison is taking too long to generate.
Unfortunately it looks like we can’t render this comparison for you right now. It might be too big, or there might be something weird with your repository.
You can try running this command locally to see the comparison on your machine:
git diff 2aa0d1f...152aaf7