Quick way to get preprocessor output???

Quick way to get preprocessor output???

[Timothy Miller]

This is in regard to the kernel message compression I'm working on.

I'm starting work on the code parser for the kernel, and I find that a 
lot of the printk format strings have macros in them.  Is there a quick 
and easy way to get preprocessor (.i) output from all kernel source files?

I'm sure I won't understand enough to hack all Makefiles, but in case I 
have to, what would be the best way of going about it?  I know about 
using -E, but I don't want to take an excessive amount of time on this.

Of course, to actually implement the text compression, all Makefiles 
will have to change anyhow to pipe preprocessor output through a program 
whose output is then run through the compiler.

Re: Quick way to get preprocessor output???

[Jörn Engel]

On Tue, 15 April 2003 11:17:02 -0400, Timothy Miller wrote:
> 
> This is in regard to the kernel message compression I'm working on.

Something tells me that fun patches are going to show up.

> I'm starting work on the code parser for the kernel, and I find that a 
> lot of the printk format strings have macros in them.  Is there a quick 
> and easy way to get preprocessor (.i) output from all kernel source files?
> 
> I'm sure I won't understand enough to hack all Makefiles, but in case I 
> have to, what would be the best way of going about it?  I know about 
> using -E, but I don't want to take an excessive amount of time on this.

Try the patch below or something similar. This was done by memory,
without testing, for 2.5. The equivalent for 2.4 would be even
simpler.

Actually, this should not work yet, the real trick would be to
generate the .o files from .c *and* the .i, so the linker stage will
find correct .o files and keep running.

In case you want to do allyesconfig, you will several hours fixing
compile problems. The resulting .config after doing all that is on my
system, you can have it if you want it.

> Of course, to actually implement the text compression, all Makefiles 
> will have to change anyhow to pipe preprocessor output through a program 
> whose output is then run through the compiler.

Not all Makefiles, just one. :)

Jörn

-- 
And spam is a useful source of entropy for /dev/random too!
-- Jasmine Strong

--- linux-2.5.67/scripts/Makefile.build~pp_output	2003-04-07 19:32:25.000000000 +0200
+++ linux-2.5.67/scripts/Makefile.build	2003-04-15 17:12:32.000000000 +0200
@@ -157,7 +157,7 @@
 	$(call if_changed_dep,cc_i_c)
 
 quiet_cmd_cc_o_c = CC $(quiet_modtag)  $@
-cmd_cc_o_c       = $(CC) $(c_flags) -c -o $@ $<
+cmd_cc_o_c       = $(CC) $(c_flags) -C -E $< > $<.i
 
 # Built-in and composite module parts