sce-update md5.txt query

[nitram]

Update checking an SCE with a missing md5.txt file completes the update check but does not flag. The md5.txt file does not appear necessary for the update check, but should it's absence be flagged to notify the user?

[Jason W]

We could do that just to make sure the sce file is not corrupt and require re-import just to be certain.

[nitram]

Thanks for response, i think the md5 integrity check during  sce-update  is a good idea. The SCE gets flagged for update if either the md5sum doesn't match the SCE.md5.txt or is missing altogether, with a note during either 'SCE update check...' scrolling or in the '(v)iew package updates' section.

[Jason W]

I will save this until after the bugfix RC being uploaded now, as it will be a somewhat involved change.

[Jason W]

Now uploaded to RC.  The 'sce-udpate -a' option does not check for this if the debinx data is current.  Of course, manual tampering with the SCE directory is not supported, but if the OLDDEBINX and NEWDEBINX don't match on 'sce-update -a' or an individual SCE is chosen for update, there must be a valid SCE.md5.txt file or a re-import will occur.  And this is echoed to /tmp/updateavailable. 

[nitram]

Deleting an md5.txt file for testing flagged it for update, despite no other library updates required, nice.

Purposely altering the md5.txt file contents (changed one digit) also flagged it, very nice.

The only feedback is could the md5 check occur first in the check cycle, as it may not be necessary to check library updates. For example, xorg-intel takes several minutes to update check all libraries but a failed md5 check takes <1 second to determine whether re-import is necessary. Checking library updates seems irrelevant at that point as the SCE will need to be re-imported fresh anyway.

[Jason W]

Makes sense, will be in the next RC I will upload shortly.

[nitram]

Just re-tested in latest RC, dCore-jessie:2016.02.19.16.43, notes:

1. Altered md5 file, update check completes almost immediately. Everything works as expected when checking the exension with  sce-update  , but not when running  sce-update -c  , provides view package updates option but it does not display any output, just exits:

Code: [Select]

sce-update -c
.
.
xorg-intel
 
Press Enter to update check above SCE(s), (q)uit to exit:
 
Checking xorg-intel.sce...
         xorg-intel update check...
 
Update check completed.
 
xorg-intel
 
Press Enter to view package updates, Ctrl-C aborts: v
tc@box:/mnt/sdb4/tce/sce$

Running  sce-update  then checking xorg-intel, check is quick and everything works as expected:

Code: [Select]

sce-update
.
.
xorg-intel
 
Press Enter to update check above SCE(s), (q)uit to exit:
 
Checking xorg-intel.sce...
         xorg-intel update check...
 
Update check completed.
 
xorg-intel
 
Press Enter to update above SCE(s), (v)iew package updates, Ctrl-C aborts: v
 
xorg-intel updates:
xorg-intel did not pass the md5sum check, needs re-importing..
 
 
Press Enter to proceed with updates, Ctrl-C aborts:

2. When booting system get extension loading error, something like 'md5sum error xorg-intel, exiting..', Some sort of error is expected, but it loads anyway despite the error. Wouldn't this be a concern if the SCE was either corrupt or maliciously altered.

[nitram]

Nevermind issue #1 above, can't read my own instructions (used ' v ' instead of Enter).

Darn scripts are so Enter friendly now

Just query #2 outstanding, thanks.

[Jason W]

Ok, should be fixed in the next RC.

[nitram]

Looked at the script change but don't understand. Running latest dCore-jessie:2016.02.21.21.07. Altered md5sum file with xorg-intel. At boot, xorg-intel loads, then other extensions load, then get xorg-intel md5sum exiting error. But xorg-intel is already loaded and i boot into graphics anyway. No change in behaviour since previous post.

[Jason W]

Should be fixed in next RC.  The loops of .dep files were the issue.   Now if the main sce or anything in it's .dep are bad md5, then neither the main sce or it's deps get loaded with that sce-load session.  However, it's deps that don't have a bad md5 can be loaded by themselves or as part of another dep file.  That is the most logical way to do it.

[nitram]

Running latest, dCore-jessie:2016.02.22.12.22. Altering the .md5.txt file contents prevented the SCE from loading at boot. Minor cosmetic issue, the error message outputs a double slash, something like sce//dillo.sce. However, removing the md5.txt file from the /sce directory allowed the extension to load normally without error at boot, so something still seems amiss, thanks.

[Jason W]

Ok, a missing md5.txt is a simple check I can put in.

[nitram]

Updated to latest RC, unattended reboot and wondered why Dillo didn't load...missing md5.txt file issue resolved, thanks