GRUB 1.90 is released

GRUB 1.90 is released

[Yoshinori K. Okuji]

I'm happy to announce the release of GRUB 1.90. This is the first of
our prereleases leading up to GRUB 2.0. We encourage you to try it, as
we have done a significant amount of work towards GRUB 2. Note that this 
release is mainly for developers who are willing to contribute to our 
project. If you need a working solution immediately, you'd better to stick to 
GRUB Legacy at the moment.

The source distribution is available from <ftp://alpha.gnu.org/gnu/grub/>.

Send any questions, ideas, bug reports to <grub-devel@gnu.org>.

Here is the relevant NEWS:

New in 1.90:

* Rename the project name PUPA to GRUB. Now this version is the
  developmental version of GRUB officially.

* The GRUB emulator ``grub-emu'' is added.

* Add support for newworld Mac. This should work with other
  PowerPC-based machines as well, if they use IEEE 1275
  (Open Firmware).

* Too many changes to describe. Look at ChangeLog for more details.

Thanks,
Okuji

Re: GRUB 1.90 is released

[Hollis Blanchard]

On Aug 7, 2005, at 12:39 PM, Yoshinori K. Okuji wrote:

> I'm happy to announce the release of GRUB 1.90.

I'm disappointed that you chose to announce GRUB 1.90 after knowingly 
breaking the PPC build.

I have fixed the build.

-Hollis

Re: GRUB 1.90 is released

[Yoshinori K. Okuji]

On Tuesday 09 August 2005 05:21, Hollis Blanchard wrote:
> On Aug 7, 2005, at 12:39 PM, Yoshinori K. Okuji wrote:
> > I'm happy to announce the release of GRUB 1.90.
>
> I'm disappointed that you chose to announce GRUB 1.90 after knowingly
> breaking the PPC build.

Do not forget that you broke it right before releasing it. Anyway, I had 
already mentioned that I would release it regardless of the status, and the 
PC support is the priority. Don't complain afterwards.

Okuji

Re: GRUB 1.90 is released

[Hollis Blanchard]

On Aug 9, 2005, at 1:32 AM, Yoshinori K. Okuji wrote:

> On Tuesday 09 August 2005 05:21, Hollis Blanchard wrote:
>> On Aug 7, 2005, at 12:39 PM, Yoshinori K. Okuji wrote:
>>> I'm happy to announce the release of GRUB 1.90.
>>
>> I'm disappointed that you chose to announce GRUB 1.90 after knowingly
>> breaking the PPC build.
>
> Do not forget that you broke it right before releasing it.

I have not forgotten, and I apologize for breaking the build. However, 
my break was accidental.

> Anyway, I had
> already mentioned that I would release it regardless of the status, 
> and the
> PC support is the priority. Don't complain afterwards.

You mentioned you wanted to release something "in the beginning of 
August." There was no rush this weekend, and I think it would have been 
polite to give people time to fix the PPC build before releasing.

-Hollis

Re: GRUB 1.90 is released

[Marco Gerards]

Hollis Blanchard <hollis@penguinppc.org> writes:

>> Anyway, I had
>> already mentioned that I would release it regardless of the status,
>> and the
>> PC support is the priority. Don't complain afterwards.
>
> You mentioned you wanted to release something "in the beginning of
> August." There was no rush this weekend, and I think it would have
> been polite to give people time to fix the PPC build before releasing.

Well, it does not really matter that much.  The most important thing
is that CVS works (right?).  The release is just something that is
formal and not really that important.

When things stabilize a bit more and when we are ready to release a
GRUB 2 version that is usable for end users we should think about how
to release.  Perhaps we need a check list or so.

Perhaps it is a good idea to send the announcement to info-gnu as
well, to attract more developers.

--
Marco

Re: GRUB 1.90 is released

[Yoshinori K. Okuji]

On Tuesday 09 August 2005 16:21, Marco Gerards wrote:
> Perhaps it is a good idea to send the announcement to info-gnu as
> well, to attract more developers.

I thought info-gnu is only for stable versions... Isn't it?

Okuji

Re: GRUB 1.90 is released

[Marco Gerards]

"Yoshinori K. Okuji" <okuji@enbug.org> writes:

> On Tuesday 09 August 2005 16:21, Marco Gerards wrote:
>> Perhaps it is a good idea to send the announcement to info-gnu as
>> well, to attract more developers.
>
> I thought info-gnu is only for stable versions... Isn't it?

It is not for test releases or releases with bug fixes.  But that is
not the case.  The list is moderated so if it is not ok, you will
notice it I guess. :)

Here is the relevant section from the maintainers guide:

http://www.gnu.org/prep/maintain/html_node/Announcements.html#Announcements

Thanks,
Marco

Re: GRUB 1.90 is released

[Yoshinori K. Okuji]

On Tuesday 09 August 2005 15:43, Hollis Blanchard wrote:
> You mentioned you wanted to release something "in the beginning of
> August." There was no rush this weekend, and I think it would have been
> polite to give people time to fix the PPC build before releasing.

I'm sorry, but I must insist my own idea about this point.

The planning has been written in the wiki. It explicitly said that the date 
would be 2005-08-07.

Besides that, the next weekend is too late. Since I have very little time in 
weekdays, if I had given it up on the 7th, I would have to postpone till the 
13th or 14th. Do you think this is still "the beginning of August"? For me, 
it is the middle of August.

I don't think my action was impolite. Because you or anyone else didn't oppose 
to my proposal of making a release "regardless of the status", I assume that 
it was an approval. Am I wrong?

I'd like to describe why I stick to the scheduling. In GRUB Legacy, I was not 
a release manager at the beginning. Gordon was the official maintainer, and I 
simply assisted him. After that, I had to manage releases myself, but I 
couldn't realize the importance of making a release regularly and quickly, 
since I had no experience. I often missed release points, because I hesitated 
to do it when I felt that something was not working or missing. That affected 
the development badly, since many people kept using old versions and didn't 
try the latest. This is one reason why GRUB Legacy has never been a stable 
version.

That gave me a lesson that it is sometimes necessary to ignore negative things 
to obtain more benefits. I do not want to make the same mistake again in GRUB 
2. If you see successful projects (at the aspect of development), most of 
them make releases very often or regularly. If you see failing projects, most 
of them do not make releases very well. I don't think I need to list up 
examples here.

As long as GRUB 2 is at a developmental phase, I'd like to keep the policy of 
making releases regularly, regardless of the status. That's because I know 
that releases can be extremely delayed, once I accept to put a delay.

At the current stage, the purpose of releases is not to provide ready-to-use 
distributions, but to get more developers interested in GRUB 2, as Marco 
pointed out. If one does not try to look at the source code only because it 
is not compilable, he/she won't take part in the development anyway, so we 
don't have to care.

I hope you would understand my point.

Okuji

Re: GRUB 1.90 is released

[Marco Gerards]

"Yoshinori K. Okuji" <okuji@enbug.org> writes:

> As long as GRUB 2 is at a developmental phase, I'd like to keep the policy of 
> making releases regularly, regardless of the status. That's because I know 
> that releases can be extremely delayed, once I accept to put a delay.

Even when you care about the status.  Okuji put the dates for the next
releases on the wiki.  If someone thinks some features are important
he just have to make sure those features are available on that date.
If you miss the release date you just have to wait for the next
release.  That is something that seams reasonable to me for everyone.

--
Marco

Re: GRUB 1.90 is released

[Hollis Blanchard]

On Aug 9, 2005, at 1:42 PM, Yoshinori K. Okuji wrote:

> On Tuesday 09 August 2005 15:43, Hollis Blanchard wrote:
>> You mentioned you wanted to release something "in the beginning of
>> August." There was no rush this weekend, and I think it would have 
>> been
>> polite to give people time to fix the PPC build before releasing.
>
> I'm sorry, but I must insist my own idea about this point.
>
> The planning has been written in the wiki. It explicitly said that the 
> date
> would be 2005-08-07.

Hmm, somehow I saw no mention of that page in the wiki. So I was quite 
surprised to go from "1.90 in August" (which to be honest I had 
forgotten about) to "I broke PPC, oh and here's GRUB 1.90!"

In the future perhaps a reminder email would be helpful. In the 
meantime I'm not too worried about what happened with 1.90; I was 
mostly surprised by the apparent abruptness.

-Hollis

Re: GRUB 1.90 is released

[Vernon Mauery]

Yoshinori K. Okuji wrote:
> I'm happy to announce the release of GRUB 1.90. This is the first of
> our prereleases leading up to GRUB 2.0. We encourage you to try it, as
> we have done a significant amount of work towards GRUB 2. Note that this 
> release is mainly for developers who are willing to contribute to our 
> project. If you need a working solution immediately, you'd better to stick to 
> GRUB Legacy at the moment.

I am seeing unaligned pointer messages when I use grub2 on /dev/hda2.  If I type 'root (hd0,2)', I get 'unaligned pointer 0x7ff94' and if I type 'insmod (hd0,2)/boot/grub/mulitboot.mod', I get 'unaligned pointer 0x7ff64'.

I downloaded the source, built it and installed it on my hard drive to test it out.  I had an unused partition that contains a fork of my live system, so I figured I could put grub2 on that.  That partition was hda3 (hd0,2).  So I ran grub-install /dev/hda3 and that went okay.  The files showed up in the right spot in /boot/grub.  Then I ran grub-setup "(hd0,2)" which also went fine.  I tried booting the image in bochs and qemu using /dev/hda as the drive for the emulated machine, chainloading from grub-legacy on (hd0) to grub2 on (hd0,2).  I get the grub prompt and all looks fine until I try to do something interesting.  Then I got the above messages.  I tried running on the actual harware rather than the emulator, but that gave the same messages, so I figured I should report them.

--Vernon

Re: GRUB 1.90 is released

[Vincent Pelletier]

-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA1

Vernon Mauery wrote:
> I am seeing unaligned pointer messages when I use grub2 on /dev/hda2.
> If I type 'root (hd0,2)', I get 'unaligned pointer 0x7ff94' and if I
> type 'insmod (hd0,2)/boot/grub/mulitboot.mod', I get 'unaligned
> pointer 0x7ff64'.

Sounds like a file system handling bug.
What FS are you using on (hd0,2) ?
By the way, (hd0,2) is hda3 not hda2.

Vincent Pelletier
-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.4.1 (GNU/Linux)

iD8DBQFC+jd+FEQoKRQyjtURAi/QAKCBJa8GBCR1YQTI8/6Plt3a1x8KUgCgiere
NOzSOH1SkQo93APSRigdMNw=
=kcAQ
-----END PGP SIGNATURE-----

	

	
		
___________________________________________________________________________ 
Appel audio GRATUIT partout dans le monde avec le nouveau Yahoo! Messenger 
Téléchargez cette version sur http://fr.messenger.yahoo.com

Re: GRUB 1.90 is released

[Vernon Mauery]

Vincent Pelletier wrote:
> Vernon Mauery wrote:
> 
>>>I am seeing unaligned pointer messages when I use grub2 on /dev/hda2.
>>>If I type 'root (hd0,2)', I get 'unaligned pointer 0x7ff94' and if I
>>>type 'insmod (hd0,2)/boot/grub/mulitboot.mod', I get 'unaligned
>>>pointer 0x7ff64'.
> 
> 
> Sounds like a file system handling bug.
> What FS are you using on (hd0,2) ?
ext3

> By the way, (hd0,2) is hda3 not hda2.
yeah, typo.  sorry.

--Vernon

Re: GRUB 1.90 is released

[Vladimir Serbinenko]

Vernon Mauery wrote:

>Vincent Pelletier wrote:
>  
>
>>Vernon Mauery wrote:
>>
>>    
>>
>>>>I am seeing unaligned pointer messages when I use grub2 on /dev/hda2.
>>>>If I type 'root (hd0,2)', I get 'unaligned pointer 0x7ff94' and if I
>>>>type 'insmod (hd0,2)/boot/grub/mulitboot.mod', I get 'unaligned
>>>>pointer 0x7ff64'.
>>>>        
>>>>
>>Sounds like a file system handling bug.
>>What FS are you using on (hd0,2) ?
>>    
>>
>ext3
>
>  
>
Try to make an fsck on this partition and repair the errors if any.
If it doesn't work could you send the characteristics of your ext3 FS?
(like size)

                                                                        
            Vladimir

Re: GRUB 1.90 is released

[Vincent Pelletier]

-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA1

Vladimir Serbinenko wrote:
> Try to make an fsck on this partition and repair the errors if any.
> If it doesn't work could you send the characteristics of your ext3 FS?
> (like size)

I'm quite sure there is a bug in ext fs handler (and probably fat), but
I don't have time to check right now...
Oh, and aren't 3-indirection level blocks unsupported ?

Vincent Pelletier
-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.4.1 (GNU/Linux)

iD8DBQFC+mczFEQoKRQyjtURAgqdAJ97ponBbZqZutynIhDb/nJPUaOMvACgqM+7
PBZ+dJGJLoK7hmPngb/P6vY=
=DYll
-----END PGP SIGNATURE-----

	

	
		
___________________________________________________________________________ 
Appel audio GRATUIT partout dans le monde avec le nouveau Yahoo! Messenger 
Téléchargez cette version sur http://fr.messenger.yahoo.com

Re: GRUB 1.90 is released

[Marco Gerards]

Vincent Pelletier <subdino2004@yahoo.fr> writes:

> Vladimir Serbinenko wrote:
>> Try to make an fsck on this partition and repair the errors if any.
>> If it doesn't work could you send the characteristics of your ext3 FS?
>> (like size)
>
> I'm quite sure there is a bug in ext fs handler (and probably fat), but
> I don't have time to check right now...

Can you tell me how to reproduce it?

> Oh, and aren't 3-indirection level blocks unsupported ?

They are, but that should not be any problem.

--
Marco

Re: GRUB 1.90 is released

[Vincent Pelletier]

[-- Warning: decoded text below may be mangled, UTF-8 assumed --]
[-- Attachment #1: Type: text/plain; charset=us-ascii, Size: 1079 bytes --]

-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA1

Marco Gerards wrote:
> Can you tell me how to reproduce it?

I had problems with one of those (damn, I can't remember now) on 720kB
floppy (erm... not really important by itself, but couldn't it be a
symptom of a deeper bug ?) but it was a while ago - when I was working
on the port and on the "play" command.
mkfs accepted without error to create the filesystem, linux was able to
mount it, store files, fsck passed. But grub displayed unaligned pointer
errors.

> They are, but that should not be any problem.

I also don't think we could find any file big enough to need 3 levels of
indirection.

Vincent Pelletier
-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.4.1 (GNU/Linux)

iD8DBQFC+y8aFEQoKRQyjtURAo++AJ9mTZHnwHYQN+Ia1y+2M8qFRC5UAwCgo590
usDL0+rKStyoX61R8kfRENs=
=ktBo
-----END PGP SIGNATURE-----

	

	
		
___________________________________________________________________________ 
Appel audio GRATUIT partout dans le monde avec le nouveau Yahoo! Messenger 
Téléchargez cette version sur http://fr.messenger.yahoo.com

Re: GRUB 1.90 is released

[Marco Gerards]

Vincent Pelletier <subdino2004@yahoo.fr> writes:

> Marco Gerards wrote:
>> Can you tell me how to reproduce it?
>
> I had problems with one of those (damn, I can't remember now) on 720kB
> floppy (erm... not really important by itself, but couldn't it be a
> symptom of a deeper bug ?) but it was a while ago - when I was working
> on the port and on the "play" command.
> mkfs accepted without error to create the filesystem, linux was able to
> mount it, store files, fsck passed. But grub displayed unaligned pointer
> errors.

Can you send me an image of this floppy?

>> They are, but that should not be any problem.
>
> I also don't think we could find any file big enough to need 3 levels of
> indirection.

It can be a problem if someone uses the loopback command.

Thanks,
Marco

Re: GRUB 1.90 is released

[Yoshinori K. Okuji]

On Thursday 11 August 2005 13:22, Marco Gerards wrote:
> > I also don't think we could find any file big enough to need 3 levels of
> > indirection.
>
> It can be a problem if someone uses the loopback command.

It's a good example.

Besides that, we must learn from the history. Many years ago, nobody could 
imagine that computers would have more than 1MB memory. So the braindead Gate 
A20 had to be invented. Likewise, nobody could imagine that hard disks would 
exceed 500MB. So LBA mode had to be invented. You can find many similar 
problems from the computer history.

The causes were sometimes technical limitations, but mostly human limitations 
that we often assume that what we see *now* are everything. Instead, we must 
always think that the future is unpredictable. If you feel that something is 
big enough in reality, it is a bad sign - because the reality changes.

In fact, some people already use quite big initrd, such as 100MB. I think it 
would be just a problem of time that people start using 1GB initrd, 
considering that the price of memory chips is decreasing quickly. Nowadays, 
even so-called embedded systems have 32MB or more memory. (So, when I see 
someone saying "necessary to optimize applications for embedded systems due 
to the resource limitations", I cannot stop laughing, since the target 
systems are always more powerful than my first "desktop computer".)

BTW, the current implementation of GRUB has the limitation of 2TB in the disk 
handling. We will have to solve this in near future.

Okuji

Re: GRUB 1.90 is released

[Vladimir Serbinenko]

Yoshinori K. Okuji wrote:

>
>Besides that, we must learn from the history. Many years ago, nobody could 
>imagine that computers would have more than 1MB memory. So the braindead Gate 
>A20 had to be invented. Likewise, nobody could imagine that hard disks would 
>exceed 500MB. So LBA mode had to be invented. You can find many similar 
>problems from the computer history.
>
>  
>
BTW. About A20: I foundan article about it:
http://www.win.tue.nl/~aeb/linux/kbd/A20.html.
Has anybody tested grub2 on i386 processor? Do you think that 2
microseconds will
invalidate the cache if processor is waiting?
                                                                        
                              Vladimir

Re: GRUB 1.90 is released

[Yoshinori K. Okuji]

On Thursday 11 August 2005 16:47, Vladimir Serbinenko wrote:
> Has anybody tested grub2 on i386 processor? Do you think that 2
> microseconds will
> invalidate the cache if processor is waiting?

Do you mean 80386? Of course, no. I don't have such an old CPU.

AFAIK, this problem may happen only in a certain model of 80386, and there is 
no way to make sure to invalidate the cache in 80386. So we have nothing to 
do with this.

BTW, why do you say 2 microseconds? Where does this come from?

Okuji

Re: GRUB 1.90 is released

[Vladimir Serbinenko]

Yoshinori K. Okuji wrote:

>
>Do you mean 80386? Of course, no. I don't have such an old CPU.
>
>  
>
But some people use old PC like a home servers. But 80386 I
suppose is not very used because it can't have more than a few
hundred Mb hard disk, which is important criteria for home server,
more important than CPU.

>AFAIK, this problem may happen only in a certain model of 80386, and there is 
>no way to make sure to invalidate the cache in 80386. So we have nothing to 
>do with this.
>  
>
Perhaps if we read a huge block of memory it will do the trick?

>BTW, why do you say 2 microseconds? Where does this come from?
>
>  
>
outb    %al, $0x80
In documentation it's written that it causes a wait of 1 microsecond

On the same page it's written that some laptops have the problem with
0x0 and 0x100000 addresses. Perhaps we should move the test address?
                                                                        
                                    Vladimir

Re: GRUB 1.90 is released

[Yoshinori K. Okuji]

On Sunday 14 August 2005 09:47, Vladimir Serbinenko wrote:
> >Do you mean 80386? Of course, no. I don't have such an old CPU.
> But some people use old PC like a home servers.

So?

> Perhaps if we read a huge block of memory it will do the trick?

Only if you prove that the current implementation does not work, I will 
consider it.

I have already investigated how other systems deal with Gate A20. For example, 
Linux does not do such a thing, but nobody has reported that it does not 
work.

> >BTW, why do you say 2 microseconds? Where does this come from?
>
> outb    %al, $0x80
> In documentation it's written that it causes a wait of 1 microsecond

Did you really read the code? GRUB tries to enable Gate A20 until it is 
enabled. What is wrong with this?

> On the same page it's written that some laptops have the problem with
> 0x0 and 0x100000 addresses. Perhaps we should move the test address?

GRUB does not use such addresses.

Okuji

Re: GRUB 1.90 is released

[Vladimir Serbinenko]

Yoshinori K. Okuji wrote:

>Only if you prove that the current implementation does not work, I will 
>consider it.
>
>I have already investigated how other systems deal with Gate A20. For example, 
>Linux does not do such a thing, but nobody has reported that it does not 
>work.
>
>  
>
Ok. after all if someone has problems with current realization he can
write a bug report. If noone reports this issue, so noone needs it

>Did you really read the code? GRUB tries to enable Gate A20 until it is 
>enabled. What is wrong with this?
>
>  
>
this outb is in checking routine. Nothing wrong

>>On the same page it's written that some laptops have the problem with
>>0x0 and 0x100000 addresses. Perhaps we should move the test address?
>>    
>>
>
>GRUB does not use such addresses.
>
>  
>
Oh sorry I just confused addl and movl

                                                                        
            Vladimir

Re: GRUB 1.90 is released

[Vernon Mauery]

Vladimir Serbinenko wrote:

> 
> Try to make an fsck on this partition and repair the errors if any.
> If it doesn't work could you send the characteristics of your ext3 FS?
> (like size)

fsck says the filesystem is clean.

bluecow:/home/vhmauery# fdisk -l

Disk /dev/hda: 80.0 GB, 80032038912 bytes
255 heads, 63 sectors/track, 9730 cylinders
Units = cylinders of 16065 * 512 = 8225280 bytes

   Device Boot      Start         End      Blocks   Id  System
/dev/hda1   *           1        1020     8193118+   7  HPFS/NTFS
/dev/hda2            1021        1033      104422+  83  Linux
/dev/hda3            1034        2338    10482412+  83  Linux
/dev/hda4            2339        9730    59376240    f  W95 Ext'd (LBA)
/dev/hda5            2339        2464     1012063+  82  Linux swap / Solaris
/dev/hda6            2465        3710    10008463+  83  Linux
/dev/hda7            3711        4318     4883728+  83  Linux
/dev/hda8            4319        6142    14651248+  8e  Linux LVM
/dev/hda9            6143        7966    14651248+  8e  Linux LVM
/dev/hda10           7967        9730    14169298+  8e  Linux LVM

--Vernon

Re: GRUB 1.90 is released

[Yoshinori K. Okuji]

On Wednesday 10 August 2005 18:06, Vernon Mauery wrote:
> I am seeing unaligned pointer messages when I use grub2 on /dev/hda2.  If I
> type 'root (hd0,2)', I get 'unaligned pointer 0x7ff94' and if I type
> 'insmod (hd0,2)/boot/grub/mulitboot.mod', I get 'unaligned pointer
> 0x7ff64'.

Thank you for your report. I think this is related to the bug which was fixed 
2 days ago. Could you try the CVS version and see if the same bug happens?

Okuji

Re: GRUB 1.90 is released

[Vernon Mauery]

Yoshinori K. Okuji wrote:
> On Wednesday 10 August 2005 18:06, Vernon Mauery wrote:
> 
>>I am seeing unaligned pointer messages when I use grub2 on /dev/hda2.  If I
>>type 'root (hd0,2)', I get 'unaligned pointer 0x7ff94' and if I type
>>'insmod (hd0,2)/boot/grub/mulitboot.mod', I get 'unaligned pointer
>>0x7ff64'.
> 
> 
> Thank you for your report. I think this is related to the bug which was fixed 
> 2 days ago. Could you try the CVS version and see if the same bug happens?
> 
the CVS version gives me the "Welcome to GRUB!" banner and then an "unaligned pointer 0x7ff58".  Instant death.

--Vernon

Re: GRUB 1.90 is released

[Yoshinori K. Okuji]

On Wednesday 10 August 2005 23:50, Vernon Mauery wrote:
> the CVS version gives me the "Welcome to GRUB!" banner and then an
> "unaligned pointer 0x7ff58".  Instant death.

Hmmh, strange. I cannot reproduce it. Can you send the file config.log? This 
might depend on the build environment.

Okuji