Please sync flashrom 0.9.1+r946-1 (universe) from Debian unstable (main)

[John Paul Adrian Glaubitz]

Binary package hint: flashrom

Hi,

the version of flashrom provided in Ubuntu lucid at the moment is quite old for flashrom standards. Usually, being a bit outdated shouldn't be a big problem. However, for flashrom, the updates in the past weeks brought many important bugfixes which should definetely make it into Lucid so that users trying flashrom for the first time won't brick their machines. Carl-Daniel Hailfinger summarizes the changes from flashrom svn r873 (the current version in Ubuntu) and r946 (todays svn version) as follows:

* r874 Add optional DMI matching to eliminate false positives in board
matching on boards with ambiguous PCI subsystem IDs
* r875 Refuse to write chips which are bigger than the addressable limit
* r877 Use partial instead of full-chip erase for better reliability
* r879 Make Dediprog SF100 support actually work for writing
* r884 Work around bogus gcc warning on newer Ubuntu
* r889 Improved memory mapping compatibility on some kernel/BIOS
combinations which caused an abort
* r892 Improve reliability of DMI matching with older mainboards
* r894 Improve reliability of SPI flashing on all Intel chipsets
* r896 Fix broken DMI mainboard matching (affects a few boards)
* r897 Fix erase of Winbond W25X{10,20,40,80}chips
* r899 Reduce the number of incorrectly detected SPI chips
* r902 Add a helpful message for recent Nvidia chipsets
* r907 Fix breakage for half a dozen chips (Pm49fl002, Pm49fl004, SST49LF160C, W39V080FA, W39V080FA_dual, 39V040C)
* r910 Fix GPO access on Intel PIIX4
* r913 Fix an incorrectly shadowed variable
* r916 Detect Laptops and warns users in that case (Important!)
* r919 Make sure only tested mainboard code is run automatically
* r922 Improve man page readability
* r930 Fix a NULL pointer bug which was triggered on a huge number of boards
* r946 Fix incorrect PCI BAR masking for memory BARs

Please see: <http://www.flashrom.org/pipermail/flashrom/2010-March/002507.html>

Fedora and Debian unstable have already updated to more recent versions, so only a re-sync with Debian unstable should be necessary to get an up-to-date version. Note that in Debian the priority of the package has been set to "high", updating this package is recommended.

Adrian

ProblemType: Bug
Architecture: amd64
Date: Thu Mar 18 14:57:09 2010
DistroRelease: Ubuntu 10.04
InstallationMedia: Ubuntu 10.04 "Lucid Lynx" - Beta amd64 (20100318)
Package: flashrom 0.9.1+r873-1
ProcEnviron:
 LANG=en_US.utf8
 SHELL=/bin/bash
ProcVersionSignature: Ubuntu 2.6.32-16.25-generic
SourcePackage: flashrom
Uname: Linux 2.6.32-16-generic x86_64

[Scott Howard (showard314)]

Formatting for a sync request:

Changelog entries since current lucid version 0.9.1+r873-1:

flashrom (0.9.1+r946-1) unstable; urgency=high

  * New upstream release.
     + Urgency high, as it fixes multiple important bugs, including such bugs
       which can potentially "brick" a user's mainboard. Details at
       http://www.flashrom.org/pipermail/flashrom/2010-March/002507.html.
     + This build should have a non-truncated manpage (Closes: #566217).
  * Standards-Version: 3.8.4 (no changes required).

 -- Uwe Hermann <email address hidden> Wed, 17 Mar 2010 09:50:05 +0100

List of upstream commits (from Adrian Glaubitz and http://www.flashrom.org/pipermail/flashrom/2010-March/002507.html)

* r874 Add optional DMI matching to eliminate false positives in board
matching on boards with ambiguous PCI subsystem IDs
* r875 Refuse to write chips which are bigger than the addressable limit
* r877 Use partial instead of full-chip erase for better reliability
* r879 Make Dediprog SF100 support actually work for writing
* r884 Work around bogus gcc warning on newer Ubuntu
* r889 Improved memory mapping compatibility on some kernel/BIOS
combinations which caused an abort
* r892 Improve reliability of DMI matching with older mainboards
* r894 Improve reliability of SPI flashing on all Intel chipsets
* r896 Fix broken DMI mainboard matching (affects a few boards)
* r897 Fix erase of Winbond W25X{10,20,40,80}chips
* r899 Reduce the number of incorrectly detected SPI chips
* r902 Add a helpful message for recent Nvidia chipsets
* r907 Fix breakage for half a dozen chips (Pm49fl002, Pm49fl004, SST49LF160C, W39V080FA, W39V080FA_dual, 39V040C)
* r910 Fix GPO access on Intel PIIX4
* r913 Fix an incorrectly shadowed variable
* r916 Detect Laptops and warns users in that case (Important!)
* r919 Make sure only tested mainboard code is run automatically
* r922 Improve man page readability
* r930 Fix a NULL pointer bug which was triggered on a huge number of boards
* r946 Fix incorrect PCI BAR masking for memory BARs

Note Debian's urgency as "high."

Built in ppa:showard314/ppa and install on Lucid.

Setting to importance high because this could brick some users machines.

[Fabrice Coutadeur (fabricesp)]

It seems to be a fixing only release. It builds and installs fine.
The only problem is that it seems that there is a missing dependency on dmidecode:
flashrom v0.9.1-r946
No coreboot table found.
sh: dmidecode: not found
...
Even without the dependency installed, it seems to run fine. Anyway, could you please check that? Old flashrom does not show this message.

Thanks,
Fabrice

[John Paul Adrian Glaubitz (glaubitz)]

Hi Fabrice,

I will check back with the flashrom guys. However, I have been told so far that the package should have "dmidecode" in the Recommends section of the control file. The explanation is that flashrom runs absolutely fine without dmidecode, however it might not detect some of the mainboards. So installing dmidecode is not a requirement but it's highly recommended as it will increase the probability that a mainboard is detected properly. We haven't fully discussed this issue yet, but if you ask me I would put "dmidecode" into the Depends section.

Adrian

[John Paul Adrian Glaubitz (glaubitz)]

Hi Fabrice,

I talked with Uwe Hermann (flashrom developer and maintainer of the Debian package), Carl-Daniel Hailfinger (project leader of flashrom) and Michael Karcher (flashrom developer) and they all agreed that dmidecode should be in the "Recommends" section of the package as I said before.

The error message:

sh: dmidecode: not found

when calling flashrom without dmidecode installed cannot be avoided, however. It's not easy to implement a platform-independent mechanism to find out whether and where dmidecode is installed on the host system. The flashrom guys said that such code would break the DOS port. But, since flashrom works perfectly despite the error message, you should use their recommended way and put dmidecode into "Recommends". When the user wants to use the full capabilities of flashrom, he/she should install dmidecode as well.

Adrian

[Fabrice Coutadeur (fabricesp)]

Hi,

Thanks for following up on that!

If I am correct, dmidecode is not in the recommends section:
Package: flashrom
Architecture: amd64 i386 kfreebsd-amd64 kfreebsd-i386 alpha hurd-i386 netbsd-alpha netbsd-i386
Depends: ${shlibs:Depends}, ${misc:Depends}
Description: Identify, read, write, erase, and verify BIOS/ROM/flash chips
...

so according to your last comment, the package should be updated. As we are close to the release date, the best solution is to add this change to the Debian version to add the recommends, and sync as soon as Lucid +1 opens.

Do you want to take care of that or I should do it?

Thanks,
Fabrice

[Carl-Daniel Hailfinger (hailfinger)]

Hi Fabrice,

does this mean that Lucid won't have flashrom 0.9.1+r964 regardless of any Debian package updates?

Maybe I should clarify the dmidecode "dependency". If dmidecode is unavailable, flashrom will spit out a warning message, but functionality will be at least on the level of any previous release or svn version. dmidecode is used for two purposes:
- Increasing match confidence for mainboards with ambiguous PCI IDs.
- Warning people against flashing if the machine is a laptop.
Both are desirable, but flashrom is designed to handle a missing dmidecode gracefully.

Admittedly the dmidecode warning message is not something people want to see, but if this depends on a speedy reaction, I'd say that the package with such a warning is a much better choice than the old version in Lynx-beta (IMHO flashing safety trumps user interface issues because a failed BIOS reflash is much more of a nightmare than any warning message).
If there is time to push an updated package to Debian and still have the result in Lynx, even better.

Regards,
Carl-Daniel

[Carl-Daniel Hailfinger (hailfinger)]

This should have been flashrom 0.9.1+r946.
Sorry for the typo.

[Fabrice Coutadeur (fabricesp)]

Hi,

That mean, and I should have made myself clearer, sorry, that instead of syncing the package from Debian, we will 'diverge' by adding this recommends, and we will sync again from Debian on next version.
So Lucid will have 0.9.1+r946: I was just asking Adrian if he wanted to prepare the update or I should take care of it, but at the end, I will upload it.

Just ot be safe, If I have no news from Adrian in 12 hours, I'll upload it myself directly.

Thanks,
Fabrice

[John Paul Adrian Glaubitz (glaubitz)]

Hi Fabrice,

I'm very sorry for my late reply, I was too tired yesterday. You want me to prepare the package for Ubuntu? That would be great. Maybe we should get in contact in IRC or so that we can discuss. In any case, I think that your proposal is just the right way to do it. I think there won't be an updated version of the package in Debian too soon, so postponing the next re-sync to lucid+1 sounds sensible to me and just diverging the package in lucid for now.

Thanks,

Adrian

[Carl-Daniel Hailfinger (hailfinger)]

Hi Fabrice,

thank you for your explanation. Such a divergence is absolutely OK, and I'm very happy that you will be uploading 0.9.1+r946 to Lucid.

Just a short note: Since flashrom only requires a small subset of dmidecode functionality, our long term plans are to integrate that functionality directly into flashrom and drop the requirement on an external dmidecode. We'll make sure to note that in the Debian package changelog once it happens.

Thanks,
Carl-Daniel

[Fabrice Coutadeur (fabricesp)]

Hi,

As we are getting closer to the release date, it will be harder to get it in, so I uploaded the new release this morning.
Thanks all for the work!

Fabrice

[Carl-Daniel Hailfinger (hailfinger)]

Thank you for uploading the new flashrom version!

Regards,
Carl-Daniel