Re: [U-Boot-Users] How to patch u-boot?
Hi
Could someone help me out with patching the u-boot.
I downloaded the source code of u-boot-1.1.6 and untared it. Then I do a
$cd u-boot-1.1.6
$svn co http://svn.openmoko.org/trunk/src/target/u-boot/patches/
those
patches are against current git head of u-boot. What gave you the
assumption
that the patches are agains 1.1.6?
Because in http://wiki.openmoko.org/wiki/Uboot, the bootloader source code section says..."The current bootloader patches can be found at http://svn.openmoko.org/trunk/src/target/u-boot/patches/. Untar the sources, apply the patch, run "make gta01bv3_config" (or gta01bv2_config, or whatever hardware revision you have), run "make". You will get a resulting "u-boot.bin" image, which you can directly flash (either using existing bootloader or sjf2410-linux) into NAND. "
So, I do not know which version to apply the patches to. How do I get the current git head of u-boot or the source code of the current bootloader for which the patches has to be applied.
Thanks Tiju Jacob
Chat on a cool, new interface. No download required. Go to http://in.messenger.yahoo.com/webmessengerpromo.php
In message 99605.88037.qm@web94612.mail.in2.yahoo.com you wrote:
Hi
Could someone help me out with patching the u-boot.
...
Stop sending such junk. Learn to send regular e-mail before poosting here again.
Best regards,
Wolfgang Denk
Tiju wrote:
Hi
[snipped unreadable original request]
Because in http://wiki.openmoko.org/wiki/Uboot, the bootloader source code section says..."The current bootloader patches can be found at http://svn.openmoko.org/trunk/src/target/u-boot/patches/. Untar the sources, apply the patch, run "make gta01bv3_config" (or gta01bv2_config, or whatever hardware revision you have), run "make". You will get a resulting "u-boot.bin" image, which you can directly flash (either using existing bootloader or sjf2410-linux) into NAND."
So, I do not know which version to apply the patches to. How do I get the current git head of u-boot or the source code of the current bootloader for which the patches has to be applied.
Thanks Tiju Jacob
Openmoko has not been an active participant on the u-boot list. They took a snapshot of u-boot and then did their own patching, but did not feed the patches back to u-boot. The result is a fork. IIRC, there has been talk about unforking openmoko, but that has not happened yet.
The bottom line is that we know little or nothing about openmoko's u-boot build methodology. You need to ask openmoko developers (openmoko lists). http://lists.openmoko.org/mailman/listinfo/
Best regards, gvb
On Mon, Jan 21, 2008 at 10:07:07AM -0500, Jerry Van Baren wrote:
Openmoko has not been an active participant on the u-boot list.
Mh, please check your list archives. I see 22 postings over the last year. Not many, but also !=- 'not been active'.
They took a snapshot of u-boot and then did their own patching, but did not feed the patches back to u-boot.
This is also not entirely true
1) we are constantly merging our patchset with mainline u-boot, at least once per week.
2) I have submitted two rounds of our patches (with about one year in between those two instances), and I have the feeling that the interest in even commenting to those patches was quite low. This is not very encouraging for further submissions.
The result is a fork. IIRC, there has been talk about unforking openmoko, but that has not happened yet.
I'd be very happy to do so. However, there are several reasons why this is extremely difficult
1) the s3c24xx chipset family support in u-boot is minimal and outdated. It only covers the s3c2400 and s3c2410, and even those only partially In order to make it support the s3c2410, s3c2440, s3c2442, I had to do some extensive re-work. Since there seems to be nobody who cares a lot about that family of chips (most products based on them, even if they run linux, don't use u-boot), I see quite a bit of reluctance to merge those patches. Furthermore, I have zero clue if there still is any living person out there who is trying to run u-boot on a s3c2400, and we certainly have no way of testing whether the new code breaks any of the old code. It has actually come to a point where I'd volunteer to maintain the s3c24xx chipset code in u-boot, if anyone was interested in that
2) Some of the changes, notably the sd/mmc driver for s3c24xx was rejected by the u-boot list, since it just does what everyone else does: no shared host controller code, just copy+paste the bits that the other sd/mmc controllers do. While I understand that there is a need for a shared sd/mmc code, putting the burden of creating such code on us is just too much. With this kind of requirement you will unlikely to see anyone merge another sd/mmc controller driver into u-boot, since everyone evades creating the generic/common code. This is not pure laziness, but inexperience with the code and lack of access to all the different hardware
3) Some code is really ugly and spread throughout the code, since there is no clean/modular way how to do this inside u-boot. One perfect example is the boot menu code that we added. I don't even bother to submit it, since I'm sure it will be rejected
The bottom line is that we know little or nothing about openmoko's u-boot build methodology. You need to ask openmoko developers (openmoko lists). http://lists.openmoko.org/mailman/listinfo/
This is entirely true. Usually those questions are raised on the (now defunct) openmoko-uboot list, that has been superseded by openmoko-kernel.
Cheers,
Harald Welte wrote:
On Mon, Jan 21, 2008 at 10:07:07AM -0500, Jerry Van Baren wrote:
Openmoko has not been an active participant on the u-boot list.
Mh, please check your list archives. I see 22 postings over the last year. Not many, but also !=- 'not been active'.
Good to know, thanks for correcting me. I've watched openmoko in the past and thought about buying one but don't have enough time in my life right now. :-(
They took a snapshot of u-boot and then did their own patching, but did not feed the patches back to u-boot.
This is also not entirely true
we are constantly merging our patchset with mainline u-boot, at least once per week.
I have submitted two rounds of our patches (with about one year in between those two instances), and I have the feeling that the interest in even commenting to those patches was quite low. This is not very encouraging for further submissions.
:-(
The result is a fork. IIRC, there has been talk about unforking openmoko, but that has not happened yet.
I'd be very happy to do so. However, there are several reasons why this is extremely difficult
- the s3c24xx chipset family support in u-boot is minimal and outdated. It only covers the s3c2400 and s3c2410, and even those only partially In order to make it support the s3c2410, s3c2440, s3c2442, I had to do some extensive re-work. Since there seems to be nobody who cares a lot about that family of chips (most products based on them, even if they run linux, don't use u-boot), I see quite a bit of reluctance to merge those patches. Furthermore, I have zero clue if there still is any living person out there who is trying to run u-boot on a s3c2400, and we certainly have no way of testing whether the new code breaks any of the old code. It has actually come to a point where I'd volunteer to maintain the s3c24xx chipset code in u-boot, if anyone was interested in that
I spend most my time in PowerPC world, so I'm pretty clueless WRT ARM-based chips. According to http://www.denx.de/wiki/UBoot/Custodians Peter Pearse is the ARM custodian, so he most likely would be the one to pester.
I would speculate the lack of comments on your patches (2 above) and the lack of anyone else(?) working with that chipset are related.
- Some of the changes, notably the sd/mmc driver for s3c24xx was rejected by the u-boot list, since it just does what everyone else does: no shared host controller code, just copy+paste the bits that the other sd/mmc controllers do. While I understand that there is a need for a shared sd/mmc code, putting the burden of creating such code on us is just too much. With this kind of requirement you will unlikely to see anyone merge another sd/mmc controller driver into u-boot, since everyone evades creating the generic/common code. This is not pure laziness, but inexperience with the code and lack of access to all the different hardware
/me == clueless
- Some code is really ugly and spread throughout the code, since there is no clean/modular way how to do this inside u-boot. One perfect example is the boot menu code that we added. I don't even bother to submit it, since I'm sure it will be rejected
:-(
The bottom line is that we know little or nothing about openmoko's u-boot build methodology. You need to ask openmoko developers (openmoko lists). http://lists.openmoko.org/mailman/listinfo/
This is entirely true. Usually those questions are raised on the (now defunct) openmoko-uboot list, that has been superseded by openmoko-kernel.
Well, at least I got one thing right. :-/
Best regards, gvb
Hi Harald,
u-boot-users-bounces@lists.sourceforge.net wrote on :
The result is a fork. IIRC, there has been talk about unforking openmoko, but that has not happened yet.
I'd be very happy to do so. However, there are several reasons why this is extremely difficult
- the s3c24xx chipset family support in u-boot is minimal and outdated. It only covers the s3c2400 and s3c2410, and even those only partially In order to make it support the s3c2410, s3c2440, s3c2442, I had to do some extensive re-work. Since there seems to be nobody who cares a lot about that family of chips (most products based on them, even if they run linux, don't use u-boot), I see quite a bit of reluctance to merge those patches. Furthermore, I have zero clue if there still is any living person out there who is trying to run u-boot on a s3c2400, and we
The trab board (with a S3C2400 CPU) is still in production. But there we use an old U-Boot version (1.1.4) and I see no need to ever update to a newer version (the board is running stable for some years now).
certainly have no way of testing whether the new code breaks any of the old code. It has actually come to a point where I'd
I would not raise any objections, if the trab board will not be supported by future U-Boot versions. I can't speak for other S3C2400 boards, but AFAIK the production of the S3C2400 has been discontinued.
Berst Regards, Martin Krause
In message 47F3F98010FF784EBEE6526EAAB078D10635DDAB@tq-mailsrv.tq-net.de you wrote:
Furthermore, I have zero clue if there still is any living person out there who is trying to run u-boot on a s3c2400, and we
The trab board (with a S3C2400 CPU) is still in production. But there we use an old U-Boot version (1.1.4) and I see no need to ever update to a newer version (the board is running stable for some years now).
Yet we still use the trab board as reference for S3C2400 when testing new U-Boot versions.
Best regards,
Wolfgang Denk
Dear Harald,
in message 20080121193613.GW706@prithivi.gnumonks.org you wrote:
- I have submitted two rounds of our patches (with about one year in between those two instances), and I have the feeling that the interest in even commenting to those patches was quite low. This is not very encouraging for further submissions.
Please don't mistake missing feedback for missing interest. There *is* interest in your patches, and care will be taken that your patches get processed as they should. Promised.
We do habe an ARM custodian problem that needs to be solved soon, but this affects all ARM contriobutions, not only yours. So please don;t take it personally.
- the s3c24xx chipset family support in u-boot is minimal and outdated.
It's outdated because nobody feeds back patches. This is a chicekn and egg situation, and if you work on this platform it's you who could solve it.
do some extensive re-work. Since there seems to be nobody who cares a lot about that family of chips (most products based on them, even if they run linux, don't use u-boot), I see quite a bit of reluctance to merge those patches. Furthermore, I have zero clue if there still
I am not reluctant. Please go on and post patches. U-Boot will not spread if we don;t use it, or if we don;t make our changes available to the public.
Pther users of such processors will not use U-Boot *because* support for these chips is "minimal and outdated".
is any living person out there who is trying to run u-boot on a s3c2400, and we certainly have no way of testing whether the new code
This is wrong. There is tens of thousands of systems running in the field.
breaks any of the old code. It has actually come to a point where I'd volunteer to maintain the s3c24xx chipset code in u-boot, if anyone was interested in that
You are welcome. Is this a serious offer? I think it might help to solve at least some parts of the ARM dilemma we're in.
- Some of the changes, notably the sd/mmc driver for s3c24xx was rejected by the u-boot list, since it just does what everyone else does: no shared host controller code, just copy+paste the bits that the other sd/mmc controllers do. While I understand that there is a need for a shared sd/mmc code, putting the burden of creating such code on us is just too much. With this kind of requirement you will unlikely to see anyone merge another sd/mmc controller driver into u-boot, since everyone evades creating the generic/common code. This is not pure laziness, but inexperience with the code and lack of access to all the different hardware
Well, someone has to go ahead...
- Some code is really ugly and spread throughout the code, since there is no clean/modular way how to do this inside u-boot. One perfect example is the boot menu code that we added. I don't even bother to submit it, since I'm sure it will be rejected
Maybe in th first version of the patch. But maybe others will add ideas how to make it less ugly and more suitable for integration, and maybe (maybe!) in the end we will have a system which is better for everybody.
Best regards,
Wolfgang Denk
Dear Wolfgang,
thanks for your prompt feedback.
On Tue, Jan 22, 2008 at 10:18:39AM +0100, Wolfgang Denk wrote:
- I have submitted two rounds of our patches (with about one year in between those two instances), and I have the feeling that the interest in even commenting to those patches was quite low. This is not very encouraging for further submissions.
Please don't mistake missing feedback for missing interest. There *is* interest in your patches, and care will be taken that your patches get processed as they should. Promised.
We do habe an ARM custodian problem that needs to be solved soon, but this affects all ARM contriobutions, not only yours. So please don;t take it personally.
I know the problem, both specifically and general to FOSS projects, as I have worked in many of them ;) So I know it isn't personally, but then also I personally don't think that my patches are more important than anyone elses...
- the s3c24xx chipset family support in u-boot is minimal and outdated.
It's outdated because nobody feeds back patches. This is a chicekn and egg situation, and if you work on this platform it's you who could solve it.
yes and no.
do some extensive re-work. Since there seems to be nobody who cares a lot about that family of chips (most products based on them, even if they run linux, don't use u-boot), I see quite a bit of reluctance to merge those patches. Furthermore, I have zero clue if there still
I am not reluctant. Please go on and post patches. U-Boot will not spread if we don;t use it, or if we don;t make our changes available to the public.
I agree that at least the core s3c24xx stuff should go mainline, no question about that.
Pther users of such processors will not use U-Boot *because* support for these chips is "minimal and outdated".
that's true. btw: Samsung is now shipping some (old, outdated) fork of u-boot with their later smdk2443 and smdk6400 boards. Obviously the code is neither clean nor suitable for submission, nor was it ever even submitted :(
But 2443 and 6400 are beyond what openmoko has done so far anyway, so this is not competing.
is any living person out there who is trying to run u-boot on a s3c2400, and we certainly have no way of testing whether the new code
This is wrong. There is tens of thousands of systems running in the field.
yes. I was thinking of 'people who develop new products and are looking for a bootloader'. The 2400 is discontinued for quite some time...
breaks any of the old code. It has actually come to a point where I'd volunteer to maintain the s3c24xx chipset code in u-boot, if anyone was interested in that
You are welcome. Is this a serious offer? I think it might help to solve at least some parts of the ARM dilemma we're in.
I'm serious about this, yes.
At least for 2410/2440/2442, I have quite extensive experience after all the hacking at OpenMoko. I'm familiar with the smdk2410, qt2410, smdk2440, qt2440, neo1973_gta01 (s3c2410), neo1973_gta02 (s3c2442b) and hxd8 (s3c2440) boards, and I can contribute board support for all those to u-boot. I'm also in contact with what Ben Dooks (Linux s3c24xx maintainer) is doing on the mainline kernel side.
- Some of the changes, notably the sd/mmc driver for s3c24xx was rejected by the u-boot list, since it just does what everyone else does: no shared host controller code, just copy+paste the bits that the other sd/mmc controllers do. While I understand that there is a need for a shared sd/mmc code, putting the burden of creating such code on us is just too much. With this kind of requirement you will unlikely to see anyone merge another sd/mmc controller driver into u-boot, since everyone evades creating the generic/common code. This is not pure laziness, but inexperience with the code and lack of access to all the different hardware
Well, someone has to go ahead...
yes, of course. But you need somebody who is voluntarily committed to doing a generic sd/mmc stack for u-boot. It's no use to try to push/force people who have only the time to write a host controlelr driver into doing it ;)
Ok. Give me a bit of time, and I'll do another round of s3c24xx/openmoko/neo1973 related patch submissions.
Cheers,
Hi Harald,
On Tuesday 22 January 2008, Harald Welte wrote:
breaks any of the old code. It has actually come to a point where I'd volunteer to maintain the s3c24xx chipset code in u-boot, if anyone was interested in that
You are welcome. Is this a serious offer? I think it might help to solve at least some parts of the ARM dilemma we're in.
I'm serious about this, yes.
At least for 2410/2440/2442, I have quite extensive experience after all the hacking at OpenMoko. I'm familiar with the smdk2410, qt2410, smdk2440, qt2440, neo1973_gta01 (s3c2410), neo1973_gta02 (s3c2442b) and hxd8 (s3c2440) boards, and I can contribute board support for all those to u-boot. I'm also in contact with what Ben Dooks (Linux s3c24xx maintainer) is doing on the mainline kernel side.
I really hope you are still willing to take over the s3c24xx U-Boot custodianshipment. This would be a great improvement. If yes, please send Wolfgang your public SSH key, so that he setup a new repository for the Samsung ARM stuff.
BTW: How should this repository be called? "u-boot-s3c24xx"? Or do you have a different suggestion?
Thanks.
Best regards, Stefan
===================================================================== DENX Software Engineering GmbH, MD: Wolfgang Denk & Detlev Zundel HRB 165235 Munich, Office: Kirchenstr.5, D-82194 Groebenzell, Germany Phone: +49-8142-66989-0 Fax: +49-8142-66989-80 Email: office@denx.de =====================================================================
Hi Stefan!
On Mon, Mar 17, 2008 at 02:54:00PM +0100, Stefan Roese wrote:
On Tuesday 22 January 2008, Harald Welte wrote:
breaks any of the old code. It has actually come to a point where I'd volunteer to maintain the s3c24xx chipset code in u-boot, if anyone was interested in that
You are welcome. Is this a serious offer? I think it might help to solve at least some parts of the ARM dilemma we're in.
I'm serious about this, yes.
At least for 2410/2440/2442, I have quite extensive experience after all the hacking at OpenMoko. I'm familiar with the smdk2410, qt2410, smdk2440, qt2440, neo1973_gta01 (s3c2410), neo1973_gta02 (s3c2442b) and hxd8 (s3c2440) boards, and I can contribute board support for all those to u-boot. I'm also in contact with what Ben Dooks (Linux s3c24xx maintainer) is doing on the mainline kernel side.
I really hope you are still willing to take over the s3c24xx U-Boot custodianshipment. This would be a great improvement. If yes, please send Wolfgang your public SSH key, so that he setup a new repository for the Samsung ARM stuff.
Yes, I am still willing to do this, and will be happy to help.
BTW: How should this repository be called? "u-boot-s3c24xx"? Or do you have a different suggestion?
That name is fine.
Please find my SSHv2 keys (home and laptop) attached.
Please also note that I'm still on holidays in Brasil until March 26th, so there will be no work before that time.
Cheers,
Hi Harald,
On Monday 17 March 2008, Harald Welte wrote:
I really hope you are still willing to take over the s3c24xx U-Boot custodianshipment. This would be a great improvement. If yes, please send Wolfgang your public SSH key, so that he setup a new repository for the Samsung ARM stuff.
Yes, I am still willing to do this, and will be happy to help.
Thanks. This is really great news.
BTW: How should this repository be called? "u-boot-s3c24xx"? Or do you have a different suggestion?
That name is fine.
Please find my SSHv2 keys (home and laptop) attached.
Please also note that I'm still on holidays in Brasil until March 26th, so there will be no work before that time.
Understood. Make sure you enjoy the vacation and the nice weather, which will be no doubt be much nicer than here in Germany. :-)
Best regards, Stefan
===================================================================== DENX Software Engineering GmbH, MD: Wolfgang Denk & Detlev Zundel HRB 165235 Munich, Office: Kirchenstr.5, D-82194 Groebenzell, Germany Phone: +49-8142-66989-0 Fax: +49-8142-66989-80 Email: office@denx.de =====================================================================
On Mon, 21 Jan 2008 20:36:13 +0100 Harald Welte laforge@openmoko.org wrote:
- I have submitted two rounds of our patches (with about one year in between those two instances), and I have the feeling that the interest in even commenting to those patches was quite low. This is not very encouraging for further submissions.
[...]
- Some of the changes, notably the sd/mmc driver for s3c24xx was rejected by the u-boot list, since it just does what everyone else does: no shared host controller code, just copy+paste the bits that the other sd/mmc controllers do.
Hey, is that the thanks I get for actually commenting at your mmc driver? I don't recall rejecting anything, nor do I have the power to do so.
While I understand that there is a need for a shared sd/mmc code, putting the burden of creating such code on us is just too much. With this kind of requirement you will unlikely to see anyone merge another sd/mmc controller driver into u-boot, since everyone evades creating the generic/common code. This is not pure laziness, but inexperience with the code and lack of access to all the different hardware
I did suggest that you move the MMC protocol definitions to a common header file. That's not a huge task, but you could have said "no, I don't want to do that right now. Maybe later." and it would have been perfectly fine. Instead, you said "will do so in the next version of the patch." But no next version of the patch ever showed up, despite quite a few comments from others as well.
Of course, it would certainly be nice with a common mmc host controller layer, but I certainly agree that placing that burden on you is unreasonable. But IMO _suggesting_ that you do it is completely different.
Btw, I find the "lack of hardware" argument a bit disturbing. Doing some cleanups really shouldn't require anyone to have access to all the affected hardware -- IMO much of the responsibility for testing that nothing breaks on real hardware lies on the custodians and other users that actually have access to the affected boards. But a few cross toolchains are definitely required to verify that it actually compiles on a handful of configurations before submitting.
Haavard
On Tue, Jan 22, 2008 at 10:32:53AM +0100, Haavard Skinnemoen wrote:
- I have submitted two rounds of our patches (with about one year in between those two instances), and I have the feeling that the interest in even commenting to those patches was quite low. This is not very encouraging for further submissions.
[...]
- Some of the changes, notably the sd/mmc driver for s3c24xx was rejected by the u-boot list, since it just does what everyone else does: no shared host controller code, just copy+paste the bits that the other sd/mmc controllers do.
Hey, is that the thanks I get for actually commenting at your mmc driver? I don't recall rejecting anything, nor do I have the power to do so.
well, sorry. But since there were no other significant comments, it was sort of the most important/influential one.
I did suggest that you move the MMC protocol definitions to a common header file. That's not a huge task, but you could have said "no, I don't want to do that right now. Maybe later." and it would have been perfectly fine. Instead, you said "will do so in the next version of the patch." But no next version of the patch ever showed up, despite quite a few comments from others as well.
well, obiously I try to implement the suggestions of the people on this list, even if it means significant additional delays. There is no point in submitting the same stuff without implementing the comments from the review.
But IMO _suggesting_ that you do it is completely different.
Well, then I say 'thanks for the suggestion'. In fact, I guess nobody would ever object to the fact that 'yes, a common layer would be better'.
Btw, I find the "lack of hardware" argument a bit disturbing. Doing some cleanups really shouldn't require anyone to have access to all the affected hardware
Well, the big issue is that most of those cleanups are deep down in the low-level head.S/start.S initialization, affecting the ordering of how PLL's are initialized, the resume-from-ram code paths, etc.
Thos can be very tricky, and especially with the level of documentation that Samsung provides about chipset bugs (close to zero), it can be very easy to break something on one chip without noticing. They have done things along the lines of re-defining the meaning of the same bits in the same registers from 'pull-up' to 'pull-down' :)
-- IMO much of the responsibility for testing that nothing breaks on real hardware lies on the custodians and other users that actually have access to the affected boards.
But a few cross toolchains are definitely required to verify that it actually compiles on a handful of configurations before submitting.
This would basically mean I had to know which toolchains other users of boards using the existing 24xx support are using.
So far, there is smdk2400 trab mpl/vcma9 sbc2410x smdk2410
Now for some of those products, there might be vendor-provided toolchains that many people use. For others, there isn't, and even if there is, quite a number of people use their own stuff, or whatever configuration of OpenEmbedded, or OpenWRT, or even something else.
Is there somewhere a list of the toolchains that the code has to compile against, including pointers on where to obtain them?
And while we're at it,
I have working (and want to add) support for: qt2410 smdk2440 neo1973/gta01 neo1973/gta02
In the future, I could potentially work on support for: some TomTom GO models (they're all 2410/244x based) smdk2443 smdk6400
Cheers,
On Tue, 22 Jan 2008 13:30:07 +0100 Harald Welte laforge@openmoko.org wrote:
On Tue, Jan 22, 2008 at 10:32:53AM +0100, Haavard Skinnemoen wrote:
Hey, is that the thanks I get for actually commenting at your mmc driver? I don't recall rejecting anything, nor do I have the power to do so.
well, sorry. But since there were no other significant comments, it was sort of the most important/influential one.
No problem. I just thought it was a bit funny that you first complained about lack of comments, and then went on to complain that the comments you _did_ get were too thorough ;-)
I did suggest that you move the MMC protocol definitions to a common header file. That's not a huge task, but you could have said "no, I don't want to do that right now. Maybe later." and it would have been perfectly fine. Instead, you said "will do so in the next version of the patch." But no next version of the patch ever showed up, despite quite a few comments from others as well.
well, obiously I try to implement the suggestions of the people on this list, even if it means significant additional delays. There is no point in submitting the same stuff without implementing the comments from the review.
Yes, I agree in general. But I also think it should be perfectly ok to say "no, sorry, that's too much work at this point", although it may start a discussion about whether it _really_ is too much work. You should never silently ignore any comments, of course.
Review comments are really just suggestions about how to improve your work further, but you shouldn't ignore them unless you have a good reason to do so, and let the reviewer and other know.
Btw, I find the "lack of hardware" argument a bit disturbing. Doing some cleanups really shouldn't require anyone to have access to all the affected hardware
Well, the big issue is that most of those cleanups are deep down in the low-level head.S/start.S initialization, affecting the ordering of how PLL's are initialized, the resume-from-ram code paths, etc.
There's mmc code down at that level...?
Thos can be very tricky, and especially with the level of documentation that Samsung provides about chipset bugs (close to zero), it can be very easy to break something on one chip without noticing. They have done things along the lines of re-defining the meaning of the same bits in the same registers from 'pull-up' to 'pull-down' :)
If you're talking about cleanups in general, then yeah, those low-level bits can be quite fragile. I think the best you can do is to test it at your own hardware and clearly state that your patch touches some tricky low-level bits of code and needs testing.
If nobody steps up to test the code, I guess we have a problem. IMHO, I'd say screw them and commit the patch anyway, after a reasonable grace period. If it turns out to break something, you'll at least find out which boards are properly maintained and which boards aren't.
-- IMO much of the responsibility for testing that nothing breaks on real hardware lies on the custodians and other users that actually have access to the affected boards.
But a few cross toolchains are definitely required to verify that it actually compiles on a handful of configurations before submitting.
This would basically mean I had to know which toolchains other users of boards using the existing 24xx support are using.
U-boot is supposed to compile with standard toolchains isn't it?
Sometimes it doesn't, but that may be an interesting piece of information on its own, which you should report to the mailing list.
Yeah, I know, avr32 currently requires a vendor-supplied toolchain, but we are working on getting it upstream (honest!) and because of this, we really don't expect people to do compile-testing with the avr32 toolchain.
Is there somewhere a list of the toolchains that the code has to compile against, including pointers on where to obtain them?
I have a standard ARM toolchain and a standard ppc toolchain, both compiled from the Ubuntu gcc-4.2 sources. Those two architectures cover a lot of configurations, so I'd expect them to uncover the vast majority of problems with a patch.
ppc is currently broken though, since nobody seems to be interested in merging my one-line fix...
Haavard
In message 20080122123007.GK706@prithivi.gnumonks.org you wrote:
Is there somewhere a list of the toolchains that the code has to compile against, including pointers on where to obtain them?
If you use ELDK 3.1.1 and ELDK 4.<latest> you will be able to cover ARM, MIPS and PowerPC; this should bring up most of the serious issues... Others will then do the rest ;-)
Best regards,
Wolfgang Denk
participants (8)
-
gvb.uboot -
Haavard Skinnemoen -
Harald Welte -
Jerry Van Baren -
Martin Krause -
Stefan Roese -
Tiju -
Wolfgang Denk