Re: Page Faults Defined

[Isaac To <kkto@csis.hku.hk>]

>>>>> "martin" == martin f krafft <madduck@debian.org> writes:

    >> > Is it protecting against segmentation faults only, or all page
    >> faults?  > And if the latter, why would you want to protect against
    >> all page > faults - after all, as I said, some are anticipated and
    >> absolutely > necessary for a machine with VM to work.
    >> 
    >> Please read the package documentation.

    martin> alright, then you're right, upstream should change the package
    martin> name.  http://www.gnu.org/software/libsigsegv/, as is quoted in
    martin> the README file is a 404 though, and yes, it might be difficult
    martin> to have them change the title. but what keeps you from giving
    martin> the .deb a better title, like libpagefault-dev?

Out of curiousity about what interface the kernel exposes to let users
changes the behaviour of the kernel page-fault handler, I've read the README
file, and I didn't get convinced that the package should be renamed.  It
handles *only* the SIGSEGV signal when the kernel is concerned, so it is
correctly named.  The "handling page fault" there in the documentation
refers to the user-visible "virtual memory page" fault, i.e., segmentation
fault as seen by the kernel.  Whether this use of the word "page fault" is
appropriate or intuitive is in the eye of beholder, but I think the long
package description should make, and had made, it clear that the program is
not just about catching those SIGSEGV caused by incorrect program, but
instead also cater for intentional SIGSEGV caused by, e.g., garbage
collectors.

BTW, I think even the word segmentation fault is a misnomer, since it has
nothing to do with segmentation.  But of course nobody is going to change
this =).

Regards,
Isaac.