Update pre-process.c to do runtime detection of gcc's internal include
paths.
This uses the same method as previously was used, it just performs the
lookup at runtime.
I do not believe this will work for cross-compiles, though, I believe
the fix will be fairly trivial. (Given that I don't cross-compile, I'm
not quite sure what the exact mechanisms to do so are.)
This should set things up to be able to provide packaged versions. My
next patch will (probably) be everything necessary to build a Debian
package.
# This is a BitKeeper generated patch for the following project:
# Project Name: TSCT - The Silly C Tokenizer
# This patch format is intended for GNU patch command version 2.5 or higher.
# This patch includes the following deltas:
# ChangeSet 1.355 -> 1.356
# Makefile 1.23 -> 1.24
# pre-process.c 1.65 -> 1.67
#
# The following is the BitKeeper ChangeSet Log
# --------------------------------------------
# 03/06/08 ryan@mythryan2.(none) 1.356
# Rework the way gcc internal includes are found to be done entirely at run-time.
# --------------------------------------------
#
diff -Nru a/Makefile b/Makefile
--- a/Makefile Sun Jun 8 21:04:32 2003
+++ b/Makefile Sun Jun 8 21:04:32 2003
@@ -32,7 +32,7 @@
expression.o: $(LIB_H)
lib.o: $(LIB_H)
parse.o: $(LIB_H)
-pre-process.o: $(LIB_H) pre-process.h
+pre-process.o: $(LIB_H)
scope.o: $(LIB_H)
show-parse.o: $(LIB_H)
symbol.o: $(LIB_H)
@@ -40,8 +40,6 @@
test-parsing.o: $(LIB_H)
tokenize.o: $(LIB_H)
-pre-process.h:
- echo "#define GCC_INTERNAL_INCLUDE \"`$(CC) -print-file-name=include`\"" > pre-process.h
clean:
- rm -f *.[oasi] core core.[0-9]* $(PROGRAMS) pre-process.h
+ rm -f *.[oasi] core core.[0-9]* $(PROGRAMS)
diff -Nru a/pre-process.c b/pre-process.c
--- a/pre-process.c Sun Jun 8 21:04:32 2003
+++ b/pre-process.c Sun Jun 8 21:04:32 2003
@@ -18,7 +18,6 @@
#include <fcntl.h>
#include <limits.h>
-#include "pre-process.h"
#include "lib.h"
#include "parse.h"
#include "token.h"
@@ -45,8 +44,7 @@
NULL,
};
-const char *gcc_includepath[] = {
- GCC_INTERNAL_INCLUDE,
+const char *gcc_includepath[INCLUDEPATHS+1] = {
NULL
};
@@ -531,6 +529,68 @@
}
return 0;
}
+
+static int init_gcc_include_path()
+{
+ char cc[128];
+ char cmd[256];
+ char *s;
+ char buffer[128];;
+ FILE *p;
+
+ s = getenv("CC");
+ if (s) {
+ strncpy(cc,s,127);
+ cc[127]='\0';
+
+ } else {
+ strcpy(cc,"gcc");
+ }
+
+ snprintf(cmd,256,"%s --print-file-name=include",cc);
+
+ p = popen(cmd,"r");
+ if (!p) {
+ perror("ERROR: failed to find gcc-include-path");
+ return 0;
+ }
+
+ /* The format of each line is simply "pathname\n".
+ * So we simply need to strip the last character off the
+ * line before storing it.
+ */
+
+ int includes = 0;
+ do {
+ s = fgets(buffer,127,p);
+ if (s) {
+ buffer[strlen(buffer)-1] = '\0';
+ gcc_includepath[includes] = strdup(buffer);
+ includes++;
+ }
+
+ } while (includes < INCLUDEPATHS - 1 && !feof(p));
+
+ gcc_includepath[includes] = NULL;
+
+ pclose(p);
+
+ return 1;
+
+
+}
+
+
+static int do_gcc_include_path(struct token *head, struct token *token, const char *filename, int flen)
+{
+ static int initialized = 0;
+ if (!initialized) {
+ if (!init_gcc_include_path())
+ return 0;
+ initialized = 1;
+ }
+ return do_include_path(gcc_includepath,head,token,filename,flen);
+}
static void do_include(int local, struct stream *stream, struct token *head, struct token *token, const char *filename)
@@ -556,7 +616,7 @@
return;
if (do_include_path(sys_includepath, head, token, filename, flen))
return;
- if (do_include_path(gcc_includepath, head, token, filename, flen))
+ if (do_gcc_include_path(head,token,filename,flen))
return;
error(token->pos, "unable to open '%s'", filename);
--
Ryan Anderson
sometimes Pug Majere
On Sun, 8 Jun 2003, Ryan Anderson wrote:
>
> This uses the same method as previously was used, it just performs the
> lookup at runtime.
I much prefer a compile-time thing.
Performance is, to me, paramount for "checker". I don't want to slow it
down, I'm hoping that some day we can just enable C=1 by default in the
kernel build (this is a _long_ time off, though, don't you all start
worrying now).
I don't see anything wrong with a compile/install time thing, that is,
after all, how gcc too works.
Linus
Followup to: <[email protected]>
By author: Linus Torvalds <[email protected]>
In newsgroup: linux.dev.kernel
>
> On Sun, 8 Jun 2003, Ryan Anderson wrote:
> >
> > This uses the same method as previously was used, it just performs the
> > lookup at runtime.
>
> I much prefer a compile-time thing.
>
> Performance is, to me, paramount for "checker". I don't want to slow it
> down, I'm hoping that some day we can just enable C=1 by default in the
> kernel build (this is a _long_ time off, though, don't you all start
> worrying now).
>
> I don't see anything wrong with a compile/install time thing, that is,
> after all, how gcc too works.
>
Both of these seem a little unnecessary. Why not pass this stuff on
the command line, and have the top-level Makefile extract the paths
into a command-line argument?
-hpa
--
<[email protected]> at work, <[email protected]> in private!
"Unix gives you enough rope to shoot yourself in the foot."
Architectures needed: ia64 m68k mips64 ppc ppc64 s390 s390x sh v850 x86-64