Noob Guestion About Make.conf - BSD

This is a discussion on Noob Guestion About Make.conf - BSD ; Confused about something - If you add -O2 -pipe will that automatically use -fomit-frame-pointer or do you need to add to make.conf (Example) CFLAGS= -O2 -pipe-fomit-frame-pointer (Will -O2 use fomit-frame-pointer by default?) I was googling and got confused about this. ...

+ Reply to Thread
Results 1 to 9 of 9

Thread: Noob Guestion About Make.conf

  1. Noob Guestion About Make.conf

    Confused about something - If you add -O2 -pipe will that automatically
    use -fomit-frame-pointer or do you need to add to make.conf

    (Example)
    CFLAGS= -O2 -pipe-fomit-frame-pointer (Will -O2 use fomit-frame-pointer
    by default?) I was googling and got confused about this. Read something
    that using -O2 would use -fomit-frame-pointer by default without having
    to add it to make.conf, read some more stuff and became confused.
    What would you suggest I do about this confusion?

    CFLAGS= -O2 -pipe-fomit-frame-pointer ....Is this *CFLAGS= -O2* the same
    as this? **CFLAGS= -O2 -pipe-fomit-frame-pointer**????
    COPTFLAGS= -O2 -pipe -fomit-frame-pointer

    Something is also breaking my programs. Is the culprit
    -ffast-math or funroll-loops?

    I can build world with ffast-math & funroll-loops and it works perfect,
    but something is screwing with my programs. (Not all programs)

    The programs will build & install but they don't work right. Firefox
    for example, it build and install fine, it works but I can't change the
    skins. What would you suggest I do about it?

    What is causing this, -fast-math or the funroll-loops?

    Thanks in advance...










  2. Re: Noob Guestion About Make.conf

    On Mon, 6 Aug 2007 01:55:07 -0400, Timmy wrote:
    > Confused about something - If you add -O2 -pipe will that
    > automatically use -fomit-frame-pointer or do you need to add to
    > make.conf


    The gcc `info documentation' is the authoritative source of
    documentation about GCC compiler options. The GCC manual says about
    this (in section 3.10 -- Options That Control Optimization):

    Not all optimizations are controlled directly by a flag. Only
    optimizations that have a flag are listed.

    % `-O'
    % `-O1'
    % [...]
    % `-O' also turns on `-fomit-frame-pointer' on machines where
    % doing so does not interfere with debugging.
    %
    % `-O2'
    % [...]
    % `-O2' turns on all optimization flags specified by `-O'. It
    % also turns on the following optimization flags:
    %
    % [...]

    So the `-fomit-frame-pointer' option MAY be turned on, but it may also
    be turned off---depending on what GCC defines as the precise meaning of
    `does not interfere with debugging'. This sort of `interference' is
    very platform-specific.

    Having written that, please make sure you read the comments near CFLAGS
    in `/usr/share/examples/etc/make.conf' and that you fully understand
    what the comments are trying to warn about:

    CFLAGS controls the compiler settings used when compiling C
    code. Note that optimization settings other than -O and -O2
    are not recommended or supported for compiling the world or
    the kernel - please revert any nonstandard optimization
    settings to "-O" or "-O2 -fno-strict-aliasing" before
    submitting bug reports without patches to the developers.

    > What would you suggest I do about this confusion?


    Nothing at all?

    If you are willing to experiment with compiler optimizations, and you
    are doing this as an exercise about compiler internals, then please go
    ahead. If you are trying to learn more about the optimizations
    supported by GCC on FreeBSD, and you are interested in helping debug any
    possible problems, then feel free to do so.

    If, on the other hand, you are merely doing this because you have read
    somewhere that it is going to magically ``make your system faster'', my
    suggestion is that you should NOT use optimizations which you cannot and
    are not willing to thoroughly test. They may cause far too many
    problems, and they may even turn out to be pessimizations instead of
    optimizations.

    > Something is also breaking my programs. Is the culprit
    > -ffast-math or funroll-loops?


    Maybe. Are you willing to experiment enough, and dig into the resulting
    object code to find out? If not, why are you using those options?

    > I can build world with ffast-math & funroll-loops and it works perfect,
    > but something is screwing with my programs. (Not all programs)


    Right. That's exactly one of the things which happen with `unsafe'
    optimizations. Some programs get screwed badly. Others don't.

    Do you really _want_ to debug the programs which are screwed and find
    out why? It may even be fun, and lead in a bugfix in either GCC or
    FreeBSD :-)

    > What is causing this, -fast-math or the funroll-loops?


    If you expect bug reports to be taken seriously, you will have to revert
    to the `standard' set of options.

    On the other hand, if you keep the extra optimizations and you *do* find
    out something about a bug, please consider reporting the bug and/or any
    fixes you find :-)

    - Giorgos


  3. Re: Noob Guestion About Make.conf

    On Mon, 06 Aug 2007 01:55:07 -0400, Timmy wrote:

    > Confused about something - If you add -O2 -pipe will that automatically
    > use -fomit-frame-pointer or do you need to add to make.conf


    On the amd64 platform (which I see you're using), my guess would be that
    no, -O2 will *not* automatically also enable -fomit-frame-pointer, as this
    *would* interfere with debugging on that platform ('info gcc' for
    details).

    You may safely use the '-pipe' switch, though. Shouldn't cause any
    problems at all, and does help to speed things up during the compile.

    If debugging is not something that's important to you, go ahead and add
    '-fomit-frame-pointer' as well. No harm done.

    > Something is also breaking my programs. Is the culprit -ffast-math or
    > funroll-loops?


    More likely -ffast-math, I would think. Options such as these should
    only be used in very special cases; they're most certainly not intended
    for general-purpose use.

    > I can build world with ffast-math & funroll-loops and it works perfect,
    > but something is screwing with my programs. (Not all programs)
    >
    > The programs will build & install but they don't work right. Firefox for
    > example, it build and install fine, it works but I can't change the
    > skins. What would you suggest I do about it?
    >
    > What is causing this, -fast-math or the funroll-loops?


    Again, the most likely culprit is -ffast-math.

    For building world/kernel, it is *highly* recommended not to mess with the
    default CFLAGS/COPTFLAGS (especially since the gcc 4.2 import, which is
    much more fussy than earlier versions, and possibly still a little buggy
    at higher optimization levels or with some of the more exotic optimization
    options). It is also strongly discouraged to report problems resulting
    from the use of non-standard compiler options with buildworld/buildkernel.

    HTH

    --
    Conrad J. Sabatier


  4. Re: Noob Guestion About Make.conf

    On Tue, 07 Aug 2007 02:24:47 GMT
    "Conrad J. Sabatier" wrote:

    > On Mon, 06 Aug 2007 01:55:07 -0400, Timmy wrote:
    >
    > > Confused about something - If you add -O2 -pipe will that
    > > automatically use -fomit-frame-pointer or do you need to add to
    > > make.conf

    >
    > On the amd64 platform (which I see you're using), my guess would be
    > that no, -O2 will *not* automatically also enable
    > -fomit-frame-pointer, as this *would* interfere with debugging on
    > that platform ('info gcc' for details).


    That's what I was wondering.. Wasn't real sure if -O2 would enable
    -fomit-frame-pointer by default.. So I turned in on in etc make.conf :-)


    > You may safely use the '-pipe' switch, though. Shouldn't cause any
    > problems at all, and does help to speed things up during the compile.


    I've always used -O2 without any problems.

    > If debugging is not something that's important to you, go ahead and
    > add '-fomit-frame-pointer' as well. No harm done.


    I rebuilt the world and a few programs from ports and haven't seen any
    problems -fomit-frame-pointer.


    > > Something is also breaking my programs. Is the culprit -ffast-math
    > > or funroll-loops?

    >
    > More likely -ffast-math, I would think. Options such as these should
    > only be used in very special cases; they're most certainly not
    > intended for general-purpose use.


    Since i've got some time on my hands I'm going to remove -ffast-math
    and just leave funroll-loops and rebuild firefox since its already in
    distfiles and see if it breaks something.


    > > I can build world with ffast-math & funroll-loops and it works
    > > perfect, but something is screwing with my programs. (Not all
    > > programs)
    > >
    > > The programs will build & install but they don't work right.
    > > Firefox for example, it build and install fine, it works but I
    > > can't change the skins. What would you suggest I do about it?
    > >
    > > What is causing this, -fast-math or the funroll-loops?

    >
    > Again, the most likely culprit is -ffast-math.
    >
    > For building world/kernel, it is *highly* recommended not to mess
    > with the default CFLAGS/COPTFLAGS (especially since the gcc 4.2
    > import, which is much more fussy than earlier versions, and possibly
    > still a little buggy at higher optimization levels or with some of
    > the more exotic optimization options). It is also strongly
    > discouraged to report problems resulting from the use of non-standard
    > compiler options with buildworld/buildkernel.


    Are you using gcc 4.2 on 6x? If so, I would also like to use gcc 4.2 on
    6.2-Stable. If you install gcc 4x it still uses gcc 3x to compile the
    world and programs.. I guess it would be a matter of pointing gcc 4x to
    the right places.. Could you tell me how to do this?

    TIA.


    > HTH
    >


  5. Re: Noob Guestion About Make.conf

    On Mon, 06 Aug 2007 09:52:18 +0300
    Giorgos Keramidas wrote:

    > On Mon, 6 Aug 2007 01:55:07 -0400, Timmy wrote:
    > > Confused about something - If you add -O2 -pipe will that
    > > automatically use -fomit-frame-pointer or do you need to add to
    > > make.conf

    >
    > The gcc `info documentation' is the authoritative source of
    > documentation about GCC compiler options. The GCC manual says about
    > this (in section 3.10 -- Options That Control Optimization):
    >
    > Not all optimizations are controlled directly by a flag. Only
    > optimizations that have a flag are listed.
    >
    > % `-O'
    > % `-O1'
    > % [...]
    > % `-O' also turns on `-fomit-frame-pointer' on machines where
    > % doing so does not interfere with debugging.
    > %
    > % `-O2'
    > % [...]
    > % `-O2' turns on all optimization flags specified by `-O'. It
    > % also turns on the following optimization flags:
    > %
    > % [...]
    >
    > So the `-fomit-frame-pointer' option MAY be turned on, but it may also
    > be turned off---depending on what GCC defines as the precise meaning
    > of `does not interfere with debugging'. This sort of `interference'
    > is very platform-specific.


    That's whats strange, it may or may not be turned on... I turned it on
    in make.conf Doesn't seem to hurt anything.


    > Having written that, please make sure you read the comments near
    > CFLAGS in `/usr/share/examples/etc/make.conf' and that you fully
    > understand what the comments are trying to warn about:


    They don't have any cool or exotic flags
    in /usr/share/examples/etc/make.conf

    > CFLAGS controls the compiler settings used when compiling C
    > code. Note that optimization settings other than -O and -O2
    > are not recommended or supported for compiling the world or
    > the kernel - please revert any nonstandard optimization
    > settings to "-O" or "-O2 -fno-strict-aliasing" before
    > submitting bug reports without patches to the developers.
    >
    > > What would you suggest I do about this confusion?

    >
    > Nothing at all?


    Don't you want to be a hacker and explore the inter-workings of your
    operating system? I like to try different things wether it's
    overclocking hardware to getting the best performance out of software.


    > If you are willing to experiment with compiler optimizations, and you
    > are doing this as an exercise about compiler internals, then please go
    > ahead. If you are trying to learn more about the optimizations
    > supported by GCC on FreeBSD, and you are interested in helping debug
    > any possible problems, then feel free to do so.


    I like to experiment but the gurus don't want to hear about your
    experience if you step outside the realm of what THEY SAY IS THE NORM.
    If you were to use max optimisations like -O3 and something breaks they
    don't want to hear about it..


    > If, on the other hand, you are merely doing this because you have read
    > somewhere that it is going to magically ``make your system faster'',
    > my suggestion is that you should NOT use optimizations which you
    > cannot and are not willing to thoroughly test. They may cause far
    > too many problems, and they may even turn out to be pessimizations
    > instead of optimizations.


    Pessimizations is something I'm lost on.. If you use funrool-loops and
    -ffast-math it will build a much larger code-base which can acutely
    slow down your programs/system, on the other-hand you may see a speed
    increase due to the optimisations used. Where is the trade off?

    > > Something is also breaking my programs. Is the culprit
    > > -ffast-math or funroll-loops?

    >
    > Maybe. Are you willing to experiment enough, and dig into the
    > resulting object code to find out? If not, why are you using those
    > options?


    I'm going to remove -ffast-math and rebuild firefox and see if its
    still broken..


    > > I can build world with ffast-math & funroll-loops and it works
    > > perfect, but something is screwing with my programs. (Not all
    > > programs)

    >
    > Right. That's exactly one of the things which happen with `unsafe'
    > optimizations. Some programs get screwed badly. Others don't.


    Would it hurt anything if you buildworld with ffastmath and
    funrool-loops and then knock those flags out when building the programs
    they break under?


    > Do you really _want_ to debug the programs which are screwed and find
    > out why? It may even be fun, and lead in a bugfix in either GCC or
    > FreeBSD :-)


    Firefox is the only program that broke. (It works, just can't install
    skins or add ons.) It seems faster.


    > > What is causing this, -fast-math or the funroll-loops?

    >
    > If you expect bug reports to be taken seriously, you will have to
    > revert to the `standard' set of options.


    Never had any problems with standard/recommended options.

    > On the other hand, if you keep the extra optimizations and you *do*
    > find out something about a bug, please consider reporting the bug
    > and/or any fixes you find :-)


    I would be glad to report it, but they don't want to hear about it..

    > - Giorgos
    >


  6. Re: Noob Guestion About Make.conf

    On Wed, 8 Aug 2007 16:26:53 -0400, Timmy wrote:
    >On Mon, 06 Aug 2007 09:52:18 +0300
    >Giorgos Keramidas wrote:
    >> CFLAGS controls the compiler settings used when compiling C
    >> code. Note that optimization settings other than -O and -O2
    >> are not recommended or supported for compiling the world or
    >> the kernel - please revert any nonstandard optimization
    >> settings to "-O" or "-O2 -fno-strict-aliasing" before
    >> submitting bug reports without patches to the developers.
    >>
    >> > What would you suggest I do about this confusion?

    >>
    >> Nothing at all?

    >
    > Don't you want to be a hacker and explore the inter-workings of your
    > operating system? I like to try different things wether it's
    > overclocking hardware to getting the best performance out of software.


    Not all the time. Some times, I just want reliable, reproducible
    results, because other things are the *real* work I want to do

    But if you like experimentation, and you want to learn more about the
    internals of the system, then please -- by all means -- go for it. The
    FreeBSD world needs all the compiler hackers it can get

    >> If you are willing to experiment with compiler optimizations, and you
    >> are doing this as an exercise about compiler internals, then please go
    >> ahead. If you are trying to learn more about the optimizations
    >> supported by GCC on FreeBSD, and you are interested in helping debug
    >> any possible problems, then feel free to do so.

    >
    > I like to experiment but the gurus don't want to hear about your
    > experience if you step outside the realm of what THEY SAY IS THE NORM.
    > If you were to use max optimisations like -O3 and something breaks they
    > don't want to hear about it..


    That's another way of saying that the 'gurus' don't want to spend their
    time tracking the bugs of _your_ setup, because they have their hands
    more than full by tracking the bugs of _their_ experiments

    >>> Something is also breaking my programs. Is the culprit -ffast-math
    >>> or funroll-loops?

    >>
    >> Maybe. Are you willing to experiment enough, and dig into the
    >> resulting object code to find out? If not, why are you using those
    >> options?

    >
    > I'm going to remove -ffast-math and rebuild firefox and see if its
    > still broken.


    Good point. If you find that it is not broken then, it may even be cool
    and hackish to find out *what* is different, see if we can find out why
    it breaks and try to find a fix!

    >> On the other hand, if you keep the extra optimizations and you *do*
    >> find out something about a bug, please consider reporting the bug
    >> and/or any fixes you find :-)

    >
    > I would be glad to report it, but they don't want to hear about it..


    That's partially true. "They", i.e. the FreeBSD team, of which I am
    also a member do not want to hear about bugs which are almost
    content-free and more or less equivalent to:

    "When I build Firefox with -O3 it breaks."

    On the other hand, if you _really_ want to help improve FreeBSD, you can
    try to trace down the precise cause of the breakage, document why it
    happens, dig into the guts of GCC itself, and post a patch which makes
    Firefox build with -O6 on FreeBSD and runs measurably faster. Then, I
    can assure you, your post will instantly get LOTS more attention.

    - Giorgos


  7. Re: Noob Guestion About Make.conf

    On Wed, 08 Aug 2007 23:57:39 +0300
    Giorgos Keramidas wrote:

    > On Wed, 8 Aug 2007 16:26:53 -0400, Timmy wrote:
    > >On Mon, 06 Aug 2007 09:52:18 +0300
    > >Giorgos Keramidas wrote:
    > >> CFLAGS controls the compiler settings used when compiling C
    > >> code. Note that optimization settings other than -O and -O2
    > >> are not recommended or supported for compiling the world or
    > >> the kernel - please revert any nonstandard optimization
    > >> settings to "-O" or "-O2 -fno-strict-aliasing" before
    > >> submitting bug reports without patches to the developers.
    > >>
    > >> > What would you suggest I do about this confusion?
    > >>
    > >> Nothing at all?

    > >
    > > Don't you want to be a hacker and explore the inter-workings of your
    > > operating system? I like to try different things wether it's
    > > overclocking hardware to getting the best performance out of
    > > software.

    >
    > Not all the time. Some times, I just want reliable, reproducible
    > results, because other things are the *real* work I want to do
    >
    > But if you like experimentation, and you want to learn more about the
    > internals of the system, then please -- by all means -- go for it.
    > The FreeBSD world needs all the compiler hackers it can get


    I would be more of an experimenter than a gcc hacker :-)

    > >> If you are willing to experiment with compiler optimizations, and
    > >> you are doing this as an exercise about compiler internals, then
    > >> please go ahead. If you are trying to learn more about the
    > >> optimizations supported by GCC on FreeBSD, and you are interested
    > >> in helping debug any possible problems, then feel free to do so.

    > >
    > > I like to experiment but the gurus don't want to hear about your
    > > experience if you step outside the realm of what THEY SAY IS THE
    > > NORM. If you were to use max optimisations like -O3 and something
    > > breaks they don't want to hear about it..

    >
    > That's another way of saying that the 'gurus' don't want to spend
    > their time tracking the bugs of _your_ setup, because they have their
    > hands more than full by tracking the bugs of _their_ experiments


    Could be..

    > >>> Something is also breaking my programs. Is the culprit -ffast-math
    > >>> or funroll-loops?
    > >>
    > >> Maybe. Are you willing to experiment enough, and dig into the
    > >> resulting object code to find out? If not, why are you using those
    > >> options?

    > >
    > > I'm going to remove -ffast-math and rebuild firefox and see if its
    > > still broken.

    >
    > Good point. If you find that it is not broken then, it may even be
    > cool and hackish to find out *what* is different, see if we can find
    > out why it breaks and try to find a fix!


    I just rebuilt Firefox using these flags:
    CFLAGS= -O2 -pipe -fomit-frame-pointer -funroll-loops

    Tested it out and it runs perfect, installed skins, addons etc.. The
    other poster was correct, it was -ffast-math causing firefox to break.


    > >> On the other hand, if you keep the extra optimizations and you *do*
    > >> find out something about a bug, please consider reporting the bug
    > >> and/or any fixes you find :-)

    > >
    > > I would be glad to report it, but they don't want to hear about it..

    >
    > That's partially true. "They", i.e. the FreeBSD team, of which I am
    > also a member


    Cool, thanks for all the hard work you guys put into what you're doing,
    your efforts gives us the best O/S on the planet. :-) Very much
    appreciated!!


    > do not want to hear about bugs which are almost
    > content-free and more or less equivalent to:
    >
    > "When I build Firefox with -O3 it breaks."


    I've narrowed down one thing in the scheme of things: funroll-loops
    break Firefox. Or at least it did on my system ;-)


    > On the other hand, if you _really_ want to help improve FreeBSD, you
    > can try to trace down the precise cause of the breakage, document why
    > it happens, dig into the guts of GCC itself, and post a patch which
    > makes Firefox build with -O6 on FreeBSD and runs measurably faster.
    > Then, I can assure you, your post will instantly get LOTS more
    > attention.


    I'm not capable of tracing this GCC breakage, at least no yet.
    I'll keep plunking around and building my calculators and little
    programs in C and C++. and experimenting with g++ and the various
    options that's available to compile said programs.

    One thing I've noticed about FBSD coming from Gentoo. When using Gentoo
    we had this whole optimisation thing going on eggdrop 24/7. Guys were
    like, I built this program with these flags, other guy is like. I've
    built my whole system using those flags + these flags etc, etc.. You
    don't really see that type of stuff on FBSD. On Gentoo, it was like who
    built what and got by with these ridiculous flags, you would hear
    stuff, yeah its a bit unstable but it works.. LMAO


    > - Giorgos
    >


  8. Re: Noob Guestion About Make.conf

    On Wed, 8 Aug 2007 19:01:30 -0400, Timmy wrote:
    >On Wed, 08 Aug 2007 23:57:39 +0300
    >Giorgos Keramidas wrote:
    >>>>> Something is also breaking my programs. Is the culprit -ffast-math
    >>>>> or funroll-loops?
    >>>>
    >>>> Maybe. Are you willing to experiment enough, and dig into the
    >>>> resulting object code to find out? If not, why are you using those
    >>>> options?
    >>>
    >>> I'm going to remove -ffast-math and rebuild firefox and see if its
    >>> still broken.

    >>
    >> Good point. If you find that it is not broken then, it may even be
    >> cool and hackish to find out *what* is different, see if we can find
    >> out why it breaks and try to find a fix!

    >
    > I just rebuilt Firefox using these flags:
    > CFLAGS= -O2 -pipe -fomit-frame-pointer -funroll-loops
    >
    > Tested it out and it runs perfect, installed skins, addons etc.. The
    > other poster was correct, it was -ffast-math causing firefox to break.


    Fantastic. That's one useful bit of information about what works and
    what doesn't. It could be helpful to see what changes when -ffast-math
    is used, and try to learn more about it, so we can understand why it
    breaks Firefox. Then, if it's still useful to use -ffast-math, we can
    fix either BSD's GCC or Firefox on BSD to work better :-)

    >> On the other hand, if you _really_ want to help improve FreeBSD, you
    >> can try to trace down the precise cause of the breakage, document why
    >> it happens, dig into the guts of GCC itself, and post a patch which
    >> makes Firefox build with -O6 on FreeBSD and runs measurably faster.
    >> Then, I can assure you, your post will instantly get LOTS more
    >> attention.

    >
    > I'm not capable of tracing this GCC breakage, at least no yet.
    > I'll keep plunking around and building my calculators and little
    > programs in C and C++. and experimenting with g++ and the various
    > options that's available to compile said programs.


    Well, life is a strange thing. You can never tell who can learn a whole
    bunch of stuff about compilers, and become a compiler guru after a few
    years


  9. Re: Noob Guestion About Make.conf

    On Wed, 08 Aug 2007 16:26:47 -0400, Timmy wrote:

    > On Tue, 07 Aug 2007 02:24:47 GMT
    > "Conrad J. Sabatier" wrote:
    >
    >> On Mon, 06 Aug 2007 01:55:07 -0400, Timmy wrote:
    >>
    >> > Confused about something - If you add -O2 -pipe will that
    >> > automatically use -fomit-frame-pointer or do you need to add to
    >> > make.conf

    >>
    >> On the amd64 platform (which I see you're using), my guess would be
    >> that no, -O2 will *not* automatically also enable -fomit-frame-pointer,
    >> as this *would* interfere with debugging on that platform ('info gcc'
    >> for details).

    >
    > That's what I was wondering.. Wasn't real sure if -O2 would enable
    > -fomit-frame-pointer by default.. So I turned in on in etc make.conf :-)
    >
    >
    >> You may safely use the '-pipe' switch, though. Shouldn't cause any
    >> problems at all, and does help to speed things up during the compile.

    >
    > I've always used -O2 without any problems.


    -O2 is, generally speaking, fine and dandy. Unfortunately, at the
    present time, it does cause buildworld breakage under CURRENT with gcc
    4.2. :-(

    >> If debugging is not something that's important to you, go ahead and add
    >> '-fomit-frame-pointer' as well. No harm done.

    >
    > I rebuilt the world and a few programs from ports and haven't seen any
    > problems -fomit-frame-pointer.


    As well you shouldn't. :-)

    >> > Something is also breaking my programs. Is the culprit -ffast-math or
    >> > funroll-loops?

    >>
    >> More likely -ffast-math, I would think. Options such as these should
    >> only be used in very special cases; they're most certainly not intended
    >> for general-purpose use.

    >
    > Since i've got some time on my hands I'm going to remove -ffast-math and
    > just leave funroll-loops and rebuild firefox since its already in
    > distfiles and see if it breaks something.


    I assume that turned out OK for you?

    >> > I can build world with ffast-math & funroll-loops and it works
    >> > perfect, but something is screwing with my programs. (Not all
    >> > programs)
    >> >
    >> > The programs will build & install but they don't work right. Firefox
    >> > for example, it build and install fine, it works but I can't change
    >> > the skins. What would you suggest I do about it?
    >> >
    >> > What is causing this, -fast-math or the funroll-loops?

    >>
    >> Again, the most likely culprit is -ffast-math.
    >>
    >> For building world/kernel, it is *highly* recommended not to mess with
    >> the default CFLAGS/COPTFLAGS (especially since the gcc 4.2 import,
    >> which is much more fussy than earlier versions, and possibly still a
    >> little buggy at higher optimization levels or with some of the more
    >> exotic optimization options). It is also strongly discouraged to
    >> report problems resulting from the use of non-standard compiler options
    >> with buildworld/buildkernel.

    >
    > Are you using gcc 4.2 on 6x? If so, I would also like to use gcc 4.2 on
    > 6.2-Stable. If you install gcc 4x it still uses gcc 3x to compile the
    > world and programs.. I guess it would be a matter of pointing gcc 4x to
    > the right places.. Could you tell me how to do this?


    Oh, I didn't realize that gcc 4.2 still hadn't been MFC'ed (merged from
    current, as we say) back to 6.x.

    I'm running 7.0-CURRENT, myself, where gcc 4.2 has been fully integrated
    now into the system, both for buildworld/buildkernel and ports.

    Not sure how good an idea it would be to try and forcibly merge 4.2 into
    6.x or STABLE on your own just yet. You may just be opening yourself up
    for a world of trouble. Best leave these things to the experts, and wait
    for it to be officially merged. :-)

    In the meantime, if you just can't resist the urge to play, you could
    install gcc 4.2 from ports and set the appropriate environment variables
    (CC, etc.) in /etc/make.conf to point to it instead of the standard gcc.
    Just do be aware that you're liable to get bit at some point. :-)

    --
    Conrad J. Sabatier


+ Reply to Thread