Win7x64 : Golden Image : partitions problems ... again

Jonathan Cool

Hi !
I ran a new debug task on a client …

[Wed Sep 12 root@fogclient ~]# blkid -po udev $part | grep FS_LABEL
blkid: The low-level probing mode requires a device

[Wed Sep 12 root@fogclient ~]# blkid                               
/dev/ram0: UUID="885cb2cc-e11f-4e88-8182-ee60a35f43d1" TYPE="ext2"
/dev/sda1: LABEL="RM-CM-)servM-CM-) au systM-CM-(me" UUID="8E4C92C44C92A70D" TYPE="ntfs" PARTUUID="c91a5cff-01"
/dev/sda2: LABEL="System" UUID="46E2941BE29410F3" TYPE="ntfs" PARTUUID="c91a5cff-02"

On the 2nd command, we can see the very strange label of /dev/sda1 … !

Edit : I deployed a classic image on o7010 and tried the same commands : same results … :

[Wed Sep 12 root@fogclient ~]# blkid
/dev/ram0: UUID="885cb2cc-e11f-4e88-8182-ee60a35f43d1" TYPE="ext2"
/dev/sda1: LABEL="RM-CM-)servM-CM-) au systM-CM-(me" UUID="A27C916A7C9139CD" TYPE="ntfs" PARTUUID="3f61e1ac-01"
/dev/sda2: LABEL="System" UUID="34C09543C0950BEC" TYPE="ntfs" PARTUUID="3f61e1ac-02"
[Wed Sep 12 root@fogclient ~]# blkid -po udev $part | grep FS_LABEL
blkid: The low-level probing mode requires a device

Sebastian Roth

@Jonathan-Cool Ok, should have listened to my intuition - those character encoding things are just a nightmare.

I think it’s wise to step back and try the route mentioned earlier. Edit /images/dev/postinitscripts/fog.postinit on your FOG server and add a single line to the end of that script so it would look like this:

#!/bin/bash
## This file serves as a starting point to call your custom pre-imaging/post init loading scripts.
## <SCRIPTNAME> should be changed to the script you're planning to use.
## Syntax of post init scripts are
#. ${postinitpath}<SCRIPTNAME>

sed -i -e "s#\\[Rr\\]\\[Ee\\]\\[Ss\\]\\[Ee\\]\\[Rr\\]\\[Vv\\]\\[Ee\\]\\[Dd\\]#RM-CM-\\\)servM-CM-\\\)#gi" /usr/share/fog/lib/funcs.sh

You know, regular expressions are just a bit tricky and those triple backslashes really look confusing. But I kind of tested this and I am fairly sure this should work if you get it all right.

Save the file and boot up your client with a debug upload task. When you get to the shell run fog and right after it said “Running post init scripts” you want to stop it for the moment (ctrl+c). Then check to see if the script did its work: grep label /usr/share/fog/lib/funcs.sh and you should see something like this:

            local label=$(getPartitionLabel "$part")
            if [[ $label =~ [Rr][Ee][Cc][Oo][Vv][Ee][Rr][Yy] ]]; then
            if [[ $label =~ RM-CM-\)servM-CM-\) || $label =~ [Rr][Éé][Ss][Éé][Rr][Vv][Éé] ]]; then
# Gets the volume group name/label.
# Get's volume labels from volume group.

If your third line looks like this I would give it a try, just start the task again using the fog command. If it does not, play with the postinitscript and try again.

Jonathan Cool

@Sebastian-Roth said in Win7x64 : Golden Image : partitions problems ... again:

Save the file and boot up your client with a debug upload task

i understand your way and your test … but if i really understood, i do the debug upload task on my VM ? Because you said “upload task” …

Sebastian Roth

@Jonathan-Cool said in Win7x64 : Golden Image : partitions problems ... again:

but if i really understood, i do the debug upload task on my VM ?

Yes, exactly what I meant!

Ohh dang, I forgot the patch for fog.upload. So please edit your /images/dev/postinitscripts/fog.postinit again and make it look like this:

#!/bin/bash
## This file serves as a starting point to call your custom pre-imaging/post init loading scripts.
## <SCRIPTNAME> should be changed to the script you're planning to use.
## Syntax of post init scripts are
#. ${postinitpath}<SCRIPTNAME>

sed -i -e "s#\\[Rr\\]\\[Ee\\]\\[Ss\\]\\[Ee\\]\\[Rr\\]\\[Vv\\]\\[Ee\\]\\[Dd\\]#RM-CM-\\\)servM-CM-\\\)#gi" /bin/fog.upload
sed -i -e "s#\\[Rr\\]\\[Ee\\]\\[Ss\\]\\[Ee\\]\\[Rr\\]\\[Vv\\]\\[Ee\\]\\[Dd\\]#RM-CM-\\\)servM-CM-\\\)#gi" /usr/share/fog/lib/funcs.sh

Hopefully I got it all right this time and it will upload the image for you in a way that you don’t have to change p1.fixed_size_partitions manually anymore.

Jonathan Cool

Hi,

I’m back after few off days …

I tried your patch on the VM, upload the new image and tried a new download … it’s didn’t work.

I tried to play with backslashes and no more sucess …

BUT, i have an other idea :

run the VM
go to the Disks Management tool
right click on the “Partition réservé au système”
rename this damn partition to “System Reserved” (<-- it is the right US/english name ?)
upload the image on FOG
try a new DL on client …

What do you think about that idea ?

Quazz

@Jonathan-Cool That should straight up work, but ofc doesn’t really help us prevent this problem in the future for new people.

Jonathan Cool

You totally right …
I will try the download with the modified name partition.
If it’s works, we can realize new tests with the “sed” command hack to permanently fix this issue for all french people.

Quazz

@Sebastian-Roth Just wondering, but would it be possible to use unicode here for the accented characters?

edit: Just tested and no…

Though I wonder, can we just use the characters directly? My small test seems to think so. Which means:

if [[ $label =~ [Rr][Ee][Ss][Ee][Rr][Vv][Ee][Dd] || $label =~ [Rr][�~Ié][Ss][�~Ié][Rr][Vv][�~Ié] ]]; then

Can be rewritten and simplified to:

if [[ $label =~ [Rr][EeÉé][Ss][EeÉé][Rr][Vv][EeÉé][Dd]? ]]; then

I’m guessing the encoding gets messed up somewhere along the road �~Ié likely simply means é. Don’t know how or when this happens/happened, but that’s my guess (I’ve seen this before trying to sort out data in different languages)

Jonathan Cool

Hm, we are close to find the issue … I can confirmed … when i renamed the system partition in my Golden Image, the reserved partition has the good size after a deploy on a o7010 …
So, the intuition was good … !

Screenshot on o7010 : https://drive.google.com/file/d/1btji1AQSUx7SmOQU5JsplMbENy6DFl1u/view?usp=sharing

Sebastian Roth

@Quazz said:

I’m guessing the encoding gets messed up somewhere along the road

Exactly what is happening. I don’t fancy trying to get the encodings right because this will break sooner than later again.

@Jonathan-Cool said:

I tried your patch on the VM, upload the new image and tried a new download … it’s didn’t work.

Can you please do another test, same way but do a debug upload and when you get to the shell run grep label /usr/share/fog/lib/funcs.sh, take a picture of the screen and post here. Just so we get an idea if the patching really works.

Jonathan Cool

Hi !

@Sebastian-Roth I ran an upload debug task on VM and did the command : https://drive.google.com/file/d/1s1CVoUuPlNNCEK7bcn6TBpclbK4hc2bK/view?usp=sharing

What do you think about the result ?

Quazz

I wanna smack myself in the head.

I think the problem is simply that the second e is accented in the test whereas it shouldn’t be.

Sebastian Roth

@Quazz said in Win7x64 : Golden Image : partitions problems ... again:

I think the problem is simply that the second e is accented in the test whereas it shouldn’t be.

Why do you think it shouldn’t be?? Reading what Jonathan posted a while ago I see two encoded e’s…

@Jonathan-Cool Looks good but… hmm?? Sorry for the forth and back here but can you please do another debug capture and run the following three commands:

grep label /usr/share/fog/lib/funcs.sh
grep label /bin/fog.upload
blkid -po udev | grep sda1

Again take a picture and post here. Just so we have it all on one screen.

Quazz

@Sebastian-Roth Yes, the first and third e should be é but, the second e should just be a plain e!

Observe

if [[ $label =~ [Rr][Ee][Ss][Ee][Rr][Vv][Ee][Dd] || $label =~ [Rr][�~Ié][Ss][�~Ié][Rr][Vv][�~Ié] ]];

The word it should look for is Réservé

It will never match under these conditions and explains why the problem remains unsolved, imo.

Jonathan Cool

Hi !

You right @Quazz … the good word is “Réservé” and not “Résérvé”.

I’m back with the results of commands (on VM Golden Img) :
https://drive.google.com/file/d/13gj87dVq2SThjy5IYdbmtiyOCZuXXtDu/view?usp=sharing

my fog.postint :

#!/bin/bash
## This file serves as a starting point to call your custom pre-imaging/post init loading scripts.
## <SCRIPTNAME> should be changed to the script you're planning to use.
## Syntax of post init scripts are
#. ${postinitpath}<SCRIPTNAME>

sed -i -e "s#\[Rr\]\[Ee\]\[Ss\]\[Ee\]\[Rr\]\[Vv\]\[Ee\]\[Dd\]#RM-CM-\\\)servM-CM-\\\)#gi" /bin/fog.upload
sed -i -e "s#\[Rr\]\[Ee\]\[Ss\]\[Ee\]\[Rr\]\[Vv\]\[Ee\]\[Dd\]#RM-CM-\\\)servM-CM-\\\)#gi" /usr/share/fog/lib/funcs.sh```

Jonathan Cool

I think, this fix is not good : https://github.com/FOGProject/fos/blob/master/Buildroot/board/FOG/FOS/rootfs_overlay/usr/share/fog/lib/funcs.sh#L493

Maybe the good fix is : line 493 :

if [[ $label =~ [Rr][Ee][Ss][Ee][Rr][Vv][Ee][Dd] || $label =~ [Rr][Éé][Ss][Ee][Rr][Vv][Éé] ]]; then

Quazz

@Jonathan-Cool I created a pull request for this, but in the mean time you can of course edit those files on your own install already. Should do the trick, I reckon.

edit: Although, looking back, part of the problem is how blkid spits out the label when using é

blkid | grep /dev/sda1

Should spit it out.

I’ve tested a bit myself and the same thing happens on my French installs.

Will need further looking into.

I also noticed I can’t type in certain characters in FOS such as é, perhaps related, will look into it later.

Sebastian Roth

@Jonathan-Cool Arghhhh, I got it wrong again. My bad!

grep Vv /usr/share/fog/lib/funcs.sh
grep Vv /bin/fog.upload
blkid -po udev /dev/sda1 | grep LABEL

Keeping my fingers crossed I get it right this time…

@Quazz Thanks for finding the tripple é’s. Didn’t notice that. My guess is that it won’t help much because the label printed is not in the same character encoding as you noticed already. Thanks for the pull request. Shall I merge? I still hope we can come up with a way to pinpoint those partitions without labels. See my detailed comment here: https://github.com/FOGProject/fogproject/issues/195 (looking for files on the partitions might be way more reliable I hope)

Jonathan Cool

Hi,

Thank you both for you precious help …

Back with the results : https://drive.google.com/file/d/1hJXQ-H7C4gZPvAUlPlVz9CzmFmBeVLK2/view?usp=sharing

@Quazz : How i can fix my funcs.sh ? i can hack it with my fog.postinit i think ? With sed tool …

Sebastian Roth

@Jonathan-Cool Holy crap (sorry)! Thanks heaps for testing both blkidcommands, the one Quazz posted and the one I did. I would have never expected those to return such different results!! As we use the blkid -po udev syntax in the scripts (ref) we are back in the race. Mind trying this script:

#!/bin/bash
## This file serves as a starting point to call your custom pre-imaging/post init loading scripts.
## <SCRIPTNAME> should be changed to the script you're planning to use.
## Syntax of post init scripts are
#. ${postinitpath}<SCRIPTNAME>

sed -i -e "s#\[Rr\]\[Ee\]\[Ss\]\[Ee\]\[Rr\]\[Vv\]\[Ee\]\[Dd\]#\[Rr\]\[Éé\]\[Ss\]\[Ee\]\[Rr\]\[Vv\]\[Éé\]#gi" /bin/fog.upload
sed -i -e "s#\[Rr\]\[Ee\]\[Ss\]\[Ee\]\[Rr\]\[Vv\]\[Ee\]\[Dd\]#\[Rr\]\[Éé\]\[Ss\]\[Ee\]\[Rr\]\[Vv\]\[Éé\]#gi" /usr/share/fog/lib/funcs.sh

Again, boot up the client in debug mode and run:

grep Vv /usr/share/fog/lib/funcs.sh
grep Vv /bin/fog.upload
blkid -po udev /dev/sda1 | grep LABEL

If those match you can go ahead and capture the image (issue command fog on the console and step through).

Win7x64 : Golden Image : partitions problems ... again

221

12.0k

17.3k

155.2k