@ kansasnoob, mc4man, ventrical and everybody else willing to *check for Startup Disk Creator bugs* in various versions of *buntu and report the result:
**Marc Deslauriers at Canonical]
On 2015-09-17 11:26 AM, Nio Wiklund wrote:
Den 2015-09-17 kl. 16:35, skrev Marc Deslauriers:
Hi!
On 2015-09-17 10:24 AM, Nio Wiklund wrote:
Hi subscribers to the ubuntu-quality mailing list,
,,,
Care to elaborate what the many bugs are? I use it every single day for testing security updates on real hardware and since bug 1279987 got fixed it's been working great for me.
...
Ideally it would also have both gtk and kde frontends, like usb-creator currently has.
Marc.
Hi Marc,
Elaborating on bugs and strange behaviour:
kansasnoob:
flaws has always been the 10 minute timeout for installing the bootloader. I'm sure I'm not the only one that multi-tasks constantly, and not just at the desk. Watching and waiting for SDC to ask if you want to install the bootloader is about like watching paint dry ............... but if you get distracted for more than 10 minutes after SDC asks about bootloader installation you have to start all over again
- Aside from SDC being badly borked one of the most annoying design
[COLOR="#aa0000"]That's a one-line fix. Is there a bug opened for that issue?[/COLOR]
- Bug numbers
https://bugs.launchpad.net/ubuntu/+source/usb-creator/+bug/1325801
But the proposed "fix" resulted in this:
https://bugs.launchpad.net/ubuntu/+source/usb-creator/+bug/1446646
My final comment on that bug was here:
https://bugs.launchpad.net/ubuntu/+source/usb-creator/+bug/1325801/comments/146
S-D-C is borked - it's useless to create live USB's of either older or newer versions of Ubuntu than the version you're running S-D-C on. And the proposed fix that made it into Vivid and got backported into Utopic resulted in S-D-C being able to only produce live USB's of the same architecture.
[COLOR="#aa0000"]Bug 1325801 is being actively worked on.
Interesting, I can't reproduce bug 1446646. I can create both amd64 and i386 bootable usb disks for trusty and vivid on my vivid amd64 laptop.
Is there something special I need to do to reproduce your issue?[/COLOR]
mc4man:
s-d-c worked ok here on 14.04.1 but fails now on installing bootloader with 14.04.2/3. s-d-c does currently work in 15.10 (trying to create 14.04.3 image)
sudodus - alias me, Nio:
There is also a bug that makes it impossible to 'erase a disk'. I don't know if it has a bug number, but it is there at least in some of the current versions.
[COLOR="#aa0000"]That's probably a udisks issue. It would help to actually have a bug filed for it.[/COLOR]
You see, there are too many bugs, that minor bug like the 'erase a disk' was not even reported. Once we thought it would be squashed when a bug
You've listed 4 bugs, two of which weren't even reported. I don't think that's too many to simply fix.
was being debugged, but it turned out to be independent of that bug, and I don't know of any bug report directed against it.
-o-
Concerning front-ends, they can be more or less independent of gtk and kde.
mkusb uses bash + zenity + pv, which are available without importing heavy stacks of gtk or kde packages.
[COLOR="#aa0000"]I do hope you're not proposing to replace s-d-c with a bash script that needs admin privileges and doesn't use udisks or policykit. That would be a non-starter. [/COLOR]
Disks belongs to gnome, I don't know if there is any corresponding built-in cloning tool in kde (not counting cp and dd).
I don't know about Unetbootin and Multisystem and the other tools, how many packages they will bring. But we can find out.
Unetbootin has versions for Windows and Mac OS too, which is an advantage.
Best regards Nio
Marc.
@ kansasnoob:
[I]1. "That's a one-line fix. Is there a bug opened for that issue?"
- "Bug 1325801 is being actively worked on.
Interesting, I can't reproduce bug 1446646. I can create both amd64 and i386 bootable usb disks for trusty and vivid on my vivid amd64 laptop.
Is there something special I need to do to reproduce your issue?" [/I] @ mc4man:
*"That's probably a udisks issue. It would help to actually have a bug filed for it."*
@ sudodus (myself) :
[I]"I do hope you're not proposing to replace s-d-c with a bash script that needs admin privileges and doesn't use udisks or policykit. That would be a non-starter."[/I]
-o-
It is not enough to *tell* the developers about the problems. We have to *file bug reports*. Yes, we know.
He first blames *udisks* for the problems in the Startup Disk Creator, and then states that a tool that should replace it must use it (udisks). And why is it a non-starter to use *sudo -H* instead of *policykit*? (I have seen that there are several problems, that are caused by the complicated policykit in more than one of the Ubuntu flavours.)
Le 2015-09-17 15:21, Nio Wiklund a écrit :
Den 2015-09-17 kl. 19:08, skrev Mathieu Trudel-Lapierre:
[...]
It seems that a fair fraction of those who have bothered to vote in the poll and / or written a detailed reply prefer Unetbootin. And I know that many of them have used it in many situations.
There have been problems with it, for example the gfxboot.c32: not a COM32R Image boot bug (corresponding to #1325801 for the SDC). But it was quickly fixed by the developer and a working version was available at the developer's PPA. Unfortunately the version of Unetbootin in the Ubuntu repository is lagging behind.
[COLOR="#aa0000"]Then please file a bug report to have it synced or whatever, as appropriate to get a recent version in the archive.[/COLOR] [...]
The remainder, when people want persistence or use the rest of the drive space for carrying data between computers, or who want a multiboot pendrive, can be managed with Unetbootin or Multisystem, but also with some other tools that 'participate' in the poll at the Ubuntu Forums. LXLE replaced the SDC with MultibootUSB in the current release. They have tested several tools before selecting it. ToriOS will soon release its first official version, and it uses mkusb.
I don't think we want to have different tools for different features. The purpose of SDC is to be an application on the default install and on the images, so that users can write a new version to USB to try it before installing. It's still satisfying this primary purpose.
FWIW, I'm not opposed to change, but I'd like to see people coming forward with something that is proven to work correctly, and that has more backing than a small poll on Forums. It's also not only my decision, but rather the decision of Ubuntu users and developers in general.
[COLOR="#aa0000"]Already looking after usb-creator, it seems like it's by far simpler to just fix the bugs. [/COLOR] So, request a session at UOS, discuss it there, and see if people are willing to do the work to support it *in Ubuntu*, looking at the bugs, etc. From there, we used to have a "defaults" session where default applications on images were "picked". If I was you I'd at least bring up your alternative on ubuntu-devel, see if there is interest. -- Mathieu Trudel-Lapierre
[I]1. "Then please file a bug report to have it synced or whatever, as appropriate to get a recent version in the archive." [/I] Yes, yes, I'll do that.
[I]2. "Already looking after usb-creator, it seems like it's by far simpler to just fix the bugs."[/I]
If it is so simple, why this never ending stream of bugs that cripple the SDC?
sudodus (nio-wiklund) :
The poor Startup Disk Creator has been suffering from many bugs for many years now, and in spite of great efforts, our developers have not managed to make it work properly. There are still several bugs and strange features.
At the Ubuntu Forums we started a discussion about replacing the Ubuntu Startup Disk Creator with another tool. See these links,
http://ubuntuforums.org/showthread.php?t=2289225
http://ubuntuforums.org/showthread.php?t=2291946
I suggest that we look for a tool that is
- easy to use for a beginner
- easy to maintain, for example more likely to work without tweaks (or with few and simple tweaks) between Ubuntu versions.
Mathieu Trudel-Lapierre (mathieu-tl) :
How about instead we just fix the bugs, given that it mostly works for people? We however can't do this unless the individual bugs are reported, and it helps a lot like people like Yu Ning provide patches too when it's very specific corner cases. There is no other tool which satisfies all of the "reliable", "easy to use" and "easy to maintain" requirements -- it's not like usb-creator is very hard to maintain, we're just dealing here with corner cases.
This particular issue here we've been at for a while, and this is simply because while I'm happy to test things and to sponsor things, usb-creator isn't the only thing I look after. Given that this isn't my top priority, I was relying on Yu Ning to complete the fixes. I did not see there was a new branch. Thankfully, we have Timo who also chimed in to help with the SRU.
As it's been mentioned previously, there are other issues which may affect behavior and make a SRU fail verification, something that we might not have thought of -- the answer to this is to iterate over these fixes, just as we've been doing now.
Can we please summarize the current state of this?
- Verification fails on Trusty and Precise.
- Does it work properly on Vivid? Presumably not because of the requirement for the architecture to be the same?
Given the same issue, it will also need to be fixed in Wily before the fix can make it to the stable releases. I'm still happy to review and sponsor proposed fixes. Mathieu Trudel-Lapierre (mathieu-tl) :
Yu Ning, please confirm that this works properly as it is in Wily and Vivid -- it seems to me like it mostly should, and we can deal with other issues in separate bugs so as not to confuse the matters worse than it already is. Changed in usb-creator (Ubuntu): status: Fix Released → said:**
That's simply false. Sure, there are some gotchas, but it largely works modulo this bug report. You're not answering the question I asked.[/COLOR]
As you may have noticed I've reset the statuses for this bug to something I think is closer to reality, and we'll close things as appropriate. We also have to consider architecture matching issues separately -- those are in bug 1446646.
[COLOR="#aa0000"]Architecture matching is unfortunately an unavoidable issue.[/COLOR] Either we run the syslinux from the system, and then you'll get the COM32R error when you boot an >utopic image written from a Trusty system, or we run the syslinux from the image about to be written and we have to match architecture or pick in-between and run the system syslinux if an i386 system tries to write a 64-bit image, and the image syslinux otherwise. Some careful application of syslinux parameters or copying additional files on the image may help too.
So I got these results from Marc Deslauriers:
- Vivid amd64 can write images from Precise and up, amd64 or i386.
I think we also should strongly suggest that users run usb-creator on a recent enough system if they don't want this bug here to happen. I'll see how release notes can be updated accordingly.
In light of all this, this bug should be Triaged/Critical for all supported releases -- we'll do another upload to -proposed to re-do the changes from Yu Ning (which appear to be fine), along with a fix for bug 1446646.
*That's simply false. Sure, there are some gotchas, but it largely works modulo this bug report. You're not answering the question I asked.*
I'll try to answer that question ...
*"Architecture matching is unfortunately an unavoidable issue."*
Yes, for the SDC. But it does not create any problems at all for tools which use dd or cp under the hood (like mkusb and Disks). And I think the other tools in our poll can manage this issue. At least Unetbootin from the developer's PPA can manage it.