Re: Exchange with Arjan

Dave,

On the current LiS-2.16.18 autoconf releases any unexported symbols are
ripped from the system map.  They don't need to be exported to be used.
Even static symbols are ripped this way.  Don't waste your breath.

--brian

On Wed, 19 Jan 2005, Dave Grothe wrote:

> 
>    Read  from  top  to  bottom.   He  is  being  his  usual  friendly and
>    cooperative self.
>    By the way, this is my favorite:
> 
>      to be
>      honest,  most  of  the  function on the big list are "wrong api"...
>      nothing in
>      the kernel using them is a reasonably good sign of that after all.
> 
>    Nothing  in  the kernel using them TODAY.  But if the need arises next
>    month or next year then there will be yet another change to the kernel
>    interface  to  bring  back the one or two symbols that suddenly become
>    useful.
>    The world view here is that the kernel source is the entire universe.
>    -- Dave
>    To: [email protected]
>    Subject:  Fwd:  [Linux-streams]  Linux  move  to  remove unused Kernel
>    symbols effects LiS
>    Hello Arjan:
>    One  of  the  LiS users sent me this.  I checked the list of undefined
>    symbols in LiS against your list and this is what I find.
>    getname
>    kill_pg
>    get_empty_filp
>    do_settimeofday
>    pci_dac_set_dma_mask
>    Of  those,  do_settimeofday  and  pci_dac_set_dma_mask  are  only used
>    because  LiS  is  providing  abstraction wrappers for those functions,
>    which  it  would simply not do if the functions were not exported from
>    the kernel.  The other three are actually used by the LiS code itself.
>    Can you retain these 5, or at least the first 3?
>    Thanks,
>    Dave
>    To: Dave Grothe <[email protected]>
>    Subject:  Re:  Fwd: [Linux-streams] Linux move to remove unused Kernel
>    symbols effects LiS
>    On Wed, Jan 19, 2005 at 02:34:08PM -0600, Dave Grothe wrote:
>    > Hello Arjan:
>    > Can you retain these 5, or at least the first 3?
>    unused is unused and takes up space in the main kernel for EVERYONE.
>    With  the cleanup we only look at mainline users realistically, and to
>    be
>    honest,  most  of  the  function  on  the  big list are "wrong api"...
>    nothing in
>    the kernel using them is a reasonably good sign of that after all.
>    (There  are  several  exceptions to that of course, which is why going
>    over
>    this  list is a manual process that has been going on for the last few
>    months
>    already)
>    To: Arjan van de Ven <[email protected]>
>    Subject:  Re:  Fwd: [Linux-streams] Linux move to remove unused Kernel
>    symbols effects LiS
>    Are  you  deleting  the functions themselves or the exporting of them?
>    Exporting is cheap.
>    -- Dave
>    To: Dave Grothe <[email protected]>
>    Subject:  Re:  Fwd: [Linux-streams] Linux move to remove unused Kernel
>    symbols effects LiS
>    On Wed, Jan 19, 2005 at 02:56:54PM -0600, Dave Grothe wrote:
>    > Are you deleting the functions themselves or the exporting of
>    > them?  Exporting is cheap.
>    if they are entirely unused in the entire kernel both.
>    if  they  are  not  used,  they  become static (which gives gcc a much
>    better
>    chance at optimizing and shrinking the code) and unexported
>    an  export  that  doesn't  result  in the function going away saves on
>    average
>    150 to 200 bytes per function. times 700......
>    unswappable  dead  wood.  Can  you justify that for ALL linux users? I
>    cannot.
>    To: Arjan van de Ven <[email protected]>
>    Subject:  Re:  Fwd: [Linux-streams] Linux move to remove unused Kernel
>    symbols effects LiS
>    But LiS is a Linux user and it uses those three functions.  So, yes, I
>    am  willing to "impose" a "penalty" of 600 bytes on ALL Linux users so
>    that LiS, and its users, can continue to function as before.
>    Sounds both cheap and friendly to me.
>    -- Dave
>    To: Dave Grothe <[email protected]>
>    Subject:  Re:  Fwd: [Linux-streams] Linux move to remove unused Kernel
>    symbols effects LiS
>    On Wed, Jan 19, 2005 at 03:10:39PM -0600, Dave Grothe wrote:
>    > But LiS is a Linux user and it uses those three functions.  So, yes,
>    I am
>    >  willing  to "impose" a "penalty" of 600 bytes on ALL Linux users so
>    that
>    > LiS, and its users, can continue to function as before.
>    >
>    > Sounds both cheap and friendly to me.
>    then I think we disagree.


-- 
Brian F. G. Bidulock    ¦ The reasonable man adapts himself to the ¦
[email protected]    ¦ world; the unreasonable one persists in  ¦
http://www.openss7.org/ ¦ trying  to adapt the  world  to himself. ¦
                        ¦ Therefore  all  progress  depends on the ¦
                        ¦ unreasonable man. -- George Bernard Shaw ¦