Marc Zyngier wrote:

quoted

quoted

quoted

quoted

In the process of moving platforms away from integrator-flash
(aka armflash), add to physmap the few features that make
armflash unique:
- optionnal probing for the AFS partition type
- init() and exit() methods, used by Integrator to control
  write access to the various onboard programmable components
Signed-off-by: Marc Zyngier <redacted>
Acked-by: Catalin Marinas <catalin.marinas@arm.com>
Cc: David Woodhouse <dwmw2@infradead.org>

[...]

quoted

quoted

quoted

quoted

diff --git a/include/linux/mtd/physmap.h b/include/linux/mtd/physmap.h
index bcfd9f7..d37cca0 100644
--- a/include/linux/mtd/physmap.h
+++ b/include/linux/mtd/physmap.h

@@ -22,6 +22,8 @@ struct map_info;

 struct physmap_flash_data {
    unsigned int            width;
+   int                     (*init)(struct platform_device *);
+   void                    (*exit)(struct platform_device *);

quoted

quoted

quoted

    How about a declaration for 'struct platform_device' beforehand? This breaks
build on other platforms at least...

quoted

quoted

Sure. Please send a patch.

quoted

uhh, what ?  you broke things, so you fix them.  i already pointed out
this breakage to you over a week ago.

Mike,

Sorry I missed your earlier comment about this. My bad.

Now, about my reply to Sergei:
      * Sergei identified the problem, and proposed a valid fix. It's
        only fair he gets credited for his work (yes, you did it before
        he did, but I didn't know it). The SOB on a patch is the proper
        way of managing the attribution.
      * Posting a patch is the quickest way for the maintainer (in this
        case David) to be able to push a fix to mainline.
      * It also helps other folks reading the ML and facing the same
        issue. They can grab the patch, apply it locally and go on with
        their own stuff until the patch is merged.
      * The person reporting the problem probably has a configuration
        showing the problem. If any of my platforms had shown that
        error, I would have fixed it long before it hit the mailing
        list.

    Actually, I'm not a primary reporter. The issue came up on the #mipslinux 
IRC channel, I then just remembered that I saw the patch set touching this file, 
and it was easy enough to complain to you after finding the corresponding mails.
    This happens to break MIPS platforms because they use warnings-as-errors 
option of gcc.

WBR, Sergei