Does "make dep" have to be run everytime kernel is built?

Does "make dep" have to be run everytime kernel is built?

Post by Ahme » Fri, 07 Jun 2002 06:38:43



Greetings,

I have a kernel for a an embedded target developed in house.  I have a
"make dep" quesiton, not particular to any arch / board.  I performed
the following steps:

$ make dep          (Step 1)
$ make clean        (Step 2)
$ make zImage       (Step 3)

Everything is fine so far.  Then

$ "changed a #define in a header file"       (Step 4)
$ make zImage                                (Step 5)

Upon which I realized the changes in the header file were not picked
up.  So I did the following

$ make clean         (Step 6)
$ make zImage        (Step 7)

Now the changes were picked up.

Q. Do I need to do a "make dep" after I do any changes to any c or h
file (before Step 5)?

Q. If the above answer is NO, why did the build process not detect the
change in the header file.  Is the dependency file creation corrupted?
 Do I need to investigate the "dep" section of the Makefile.

Q. How do I go about debugging this issue.

Thanks in advance,
Ahmed.

 
 
 

Does "make dep" have to be run everytime kernel is built?

Post by rain » Sat, 08 Jun 2002 21:38:51



> Q. Do I need to do a "make dep" after I do any changes to any c or h
> file (before Step 5)?

The current kernel's Makefile is quite broken according to LKML.  Personally,
I mostly apply patches only, so I always `make mrproper' (which is a full
clean, so save config first) after patching the kernel.  I haven't modified
any of the headers, although I once played with sched.c, and of course the
Makefile does the right thing.

I think you have 3 choices:
1. Install the new kbuild 2.5 patch by Keith.  It should not affect actual
   code, but it might be a bit difficult to install.
2. Rebuild the whole tree after changing a header --- not viable, I think.
3. `touch' the affected files manually after changing headers.  If it is an
   arch-specific header or a driver, probably it is only used by files in
   the relevent directories, so you may touch them all.  If it is something
   like mm.h or fs.h, most of the kernel had to be rebuilt anyway, so just
   do a full rebuild.

Quote:> Q. If the above answer is NO, why did the build process not detect the
> change in the header file.  Is the dependency file creation corrupted?
>  Do I need to investigate the "dep" section of the Makefile.

> Q. How do I go about debugging this issue.

Since this seems to have been done in the new kbuild patch, it is not very
advisable to reinvent the wheel.  Just try the patch, or live with the current
conditions like most kernel developers.
Quote:

> Thanks in advance,
> Ahmed.


 
 
 

1. I am having problems with "fvwm2" or Xwindows

First, I am a newbie on linux so be patient.
Actually, i am not a real newbie because I no how to navigate, run programs,
edit them and all the basic stuff.

For my problem well, It's just that when i finished installing RedHat 5.1
I ran all the configuration stuff then typed "startx" and It worked.
But when i restarted linux (a few hours later) I tried it and at the
begining it was normal (mouse showing up) but a few second later It shuted
down it self and sometimes I got some echo on my text (ex: "startx" wood be
"stttaaxx" or something like that and also at the top of the screen it said
"400x300 screen test unable to sync h. at 67 Hz operation canceled" or
something like that and it said it a few other times with difrent sync.
So if you think you can help me HELP ME!


Thanks in advance.

Julian Choquette

2. Touchscreen control

3. Am I the only one who can't "make build" 2.9?

4. Linux & Netscape !!!

5. strange output of "who am I" & "last" on solaris 9

6. Can you get a PowerDisplay 19 to work with a PC

7. strange output of "who am i" & "last" in Solaris 9

8. Linux Distributions

9. GETSERVBYNAME()????????????????????"""""""""""""

10. newly built kernel "loads" but doesn't "uncompress"

11. """"""""My SoundBlast 16 pnp isn't up yet""""""""""""

12. I am with the following error, when i am running lilo...

13. Having "real" and "private" IPs work together on one hub