-BDIRECT, anyone seen this?

Message

neuron · Post by **neuron** » Fri Nov 04, 2005 2:38 pm

Somehow this post:
http://sourceware.org/ml/binutils/2005-10/msg00436.html

Has gone without any discussion, and with very little publisity at all, but personally I find it extreamly interesting. The counter argument is that prelink already does something like this, but in my opinion prelink is a dirty hack (set up manually, or if when set up by the distro runs a task which uses all my resources on an interval, no thank you), and this solution seems relativly clean.

Anyone else noticed this thread at all?

nxsty · Post by **nxsty** » Sat Nov 05, 2005 9:21 am

Sounds nice. But the toolchain dictator didn't like it so it'll probably never be included.

IWBCMAN · Post by **IWBCMAN** » Sat Nov 05, 2005 9:56 am

yup, Ulrich Drepper, definitely deserves the name toolchain dictator. It is just one of those things...if you do the hard work, which others won't/don't do, you have the right to be an asshole-if others don't like it-tough luck....I am sure that he has his reasons(and probably good ones too) for rejecting Michael Meeks suggested patch..but I am equally sure that he couldn't be bothered to actually discuss why such an approach as suggested is a wrong way to solve the percieved problem. The kind of terseness in his comment - his non-chalant way of utterly dismissing Meeks is enough to bring ones blood to a boil- but it is his perogative, even if, in so doing, he is a callous asshole. But I trully do value the work of that asshole, I certainly can't do it, and I know of no one who could do it better, so who am I to complain....It is a sahme that some of the best coders, some of the ones who do the best work, that make our lives so much better are such insufferable assholes-but such is life and we would be poorer without them

playfool · Post by **playfool** » Sat Nov 05, 2005 10:24 am

Ulrich Drepper is one of the single most talented programmers in the community, if he rejects something he has his reasons.

nxsty · Post by **nxsty** » Sat Nov 05, 2005 10:55 am

playfool wrote:Ulrich Drepper is one of the single most talented programmers in the community, if he rejects something he has his reasons.

He is very talented indeed but I've seen him reject other seemingly nice patches with similar useless motivations. If he has good reasons, why don't he just say them?

Martigen · Post by **Martigen** » Sat Nov 05, 2005 11:02 am

nxsty, can you include Meeks' patch in your next patched glibc? It sounds like a clean patch, and easily tested/ignored if people so choose. Prelink is a fudge, and doesn't work completely. If Ulrich doesn't want to give it a go, doesn't mean we can't (and even, perhaps, report results).

M

nxsty · Post by **nxsty** » Sat Nov 05, 2005 11:42 am

Martigen wrote:nxsty, can you include Meeks' patch in your next patched glibc? It sounds like a clean patch, and easily tested/ignored if people so choose. Prelink is a fudge, and doesn't work completely. If Ulrich doesn't want to give it a go, doesn't mean we can't (and even, perhaps, report results)

I'll probably not. The patch only seems lile an uncompletet draft yet. And besides you'll have to patch both binutils and glibc and recompile everything with -Bdirect to make it effective.

There where another patch sent to glibc-alpha that did some kind of readahead to improve app startup time but it was also ignored like this one.
http://marc.theaimsgroup.com/?l=glibc-a ... 720762&w=2

playfool · Post by **playfool** » Sat Nov 05, 2005 1:33 pm

nxsty wrote:
playfool wrote:Ulrich Drepper is one of the single most talented programmers in the community, if he rejects something he has his reasons.
He is very talented indeed but I've seen him reject other seemingly nice patches with similar useless motivations. If he has good reasons, why don't he just say them?

Because rejecting crap patches sours people, doing it time after time tends to turn people into Al Viro (if you think Drepper is bad read some Al Viro replies to kernel patches) - I'm not saying it's okay but Drepper has proven his skill time and time again, he knows what he's doing - everyone is free to clean up the patch and test it in their own branch, something I'm sure Novell will consider, if it proves itself there the glibc maintainers will surely be convinced to include it.

enderandrew · Post by **enderandrew** » Fri Nov 11, 2005 9:09 am

Has anyone tried this patch?

playfool · Post by **playfool** » Wed Nov 30, 2005 2:15 pm

And Micheal Meeks just submitted this to our bugzilla:

http://bugs.gentoo.org/show_bug.cgi?id=114008

So now you to can enjoy -Bdirect goodness.

neuron · Post by **neuron** » Wed Nov 30, 2005 4:20 pm

very nice, thanks for the heads up

wangxiaohu · Post by **wangxiaohu** » Mon Dec 05, 2005 5:33 pm

playfool wrote:And Micheal Meeks just submitted this to our bugzilla:

http://bugs.gentoo.org/show_bug.cgi?id=114008

So now you to can enjoy -Bdirect goodness.

Can I just simply add -Bdirect to LD_FLAGS in make.conf? As said in the comments, it brings troubles to glibc, so does glibc strip it correctly now?

predatorfreak · Post by **predatorfreak** » Tue Dec 06, 2005 3:35 am

wangxiaohu wrote:
playfool wrote:And Micheal Meeks just submitted this to our bugzilla:

http://bugs.gentoo.org/show_bug.cgi?id=114008

So now you to can enjoy -Bdirect goodness.
Can I just simply add -Bdirect to LD_FLAGS in make.conf? As said in the comments, it brings troubles to glibc, so does glibc strip it correctly now?

It's useless until glibc supports it, right now only binutils supports it.

nxsty · Post by **nxsty** » Tue Dec 06, 2005 2:23 pm

predatorfreak wrote:
wangxiaohu wrote:
playfool wrote:And Micheal Meeks just submitted this to our bugzilla:

http://bugs.gentoo.org/show_bug.cgi?id=114008

So now you to can enjoy -Bdirect goodness.
Can I just simply add -Bdirect to LD_FLAGS in make.conf? As said in the comments, it brings troubles to glibc, so does glibc strip it correctly now?
It's useless until glibc supports it, right now only binutils supports it.

I added the patch to my glibc overlay if you want to try it. The flag breaks glibc so it's filtered from the build but it might break other builds too. There is only one way to find out.