diff options
author | David Robillard <d@drobilla.net> | 2019-04-21 23:05:06 +0200 |
---|---|---|
committer | David Robillard <d@drobilla.net> | 2019-04-21 23:05:06 +0200 |
commit | 202873cf9bfdeb669365a6d76e54ad657c550fe2 (patch) | |
tree | 79db4dee998f1ef093c3e81f1d0ea708901ab96c /waflib/extras/c_dumbpreproc.py | |
parent | de6492738adf1794bf5fa39c1fe1ebbd167727ac (diff) | |
download | sratom-202873cf9bfdeb669365a6d76e54ad657c550fe2.tar.gz sratom-202873cf9bfdeb669365a6d76e54ad657c550fe2.tar.bz2 sratom-202873cf9bfdeb669365a6d76e54ad657c550fe2.zip |
Switch to using a submodule for autowaf
Diffstat (limited to 'waflib/extras/c_dumbpreproc.py')
m--------- | waflib | 0 | ||||
-rw-r--r-- | waflib/extras/c_dumbpreproc.py | 72 |
2 files changed, 0 insertions, 72 deletions
diff --git a/waflib b/waflib new file mode 160000 +Subproject 2314e236ca6e7d94a26c3c17091da0f25f5867f diff --git a/waflib/extras/c_dumbpreproc.py b/waflib/extras/c_dumbpreproc.py deleted file mode 100644 index ce9e1a4..0000000 --- a/waflib/extras/c_dumbpreproc.py +++ /dev/null @@ -1,72 +0,0 @@ -#!/usr/bin/env python -# encoding: utf-8 -# Thomas Nagy, 2006-2010 (ita) - -""" -Dumb C/C++ preprocessor for finding dependencies - -It will look at all include files it can find after removing the comments, so the following -will always add the dependency on both "a.h" and "b.h":: - - #include "a.h" - #ifdef B - #include "b.h" - #endif - int main() { - return 0; - } - -To use:: - - def configure(conf): - conf.load('compiler_c') - conf.load('c_dumbpreproc') -""" - -import re -from waflib.Tools import c_preproc - -re_inc = re.compile( - '^[ \t]*(#|%:)[ \t]*(include)[ \t]*[<"](.*)[>"]\r*$', - re.IGNORECASE | re.MULTILINE) - -def lines_includes(node): - code = node.read() - if c_preproc.use_trigraphs: - for (a, b) in c_preproc.trig_def: - code = code.split(a).join(b) - code = c_preproc.re_nl.sub('', code) - code = c_preproc.re_cpp.sub(c_preproc.repl, code) - return [(m.group(2), m.group(3)) for m in re.finditer(re_inc, code)] - -parser = c_preproc.c_parser -class dumb_parser(parser): - def addlines(self, node): - if node in self.nodes[:-1]: - return - self.currentnode_stack.append(node.parent) - - # Avoid reading the same files again - try: - lines = self.parse_cache[node] - except KeyError: - lines = self.parse_cache[node] = lines_includes(node) - - self.lines = lines + [(c_preproc.POPFILE, '')] + self.lines - - def start(self, node, env): - try: - self.parse_cache = node.ctx.parse_cache - except AttributeError: - self.parse_cache = node.ctx.parse_cache = {} - - self.addlines(node) - while self.lines: - (x, y) = self.lines.pop(0) - if x == c_preproc.POPFILE: - self.currentnode_stack.pop() - continue - self.tryfind(y) - -c_preproc.c_parser = dumb_parser - |