I'm inclined to think that a failed to remount error is at least
potentially more useful when it doesn't include SD card not being present
at all errors.
On Feb 22, 2016 12:46 PM, "Hzj_jie" notifications@github.com wrote:

I surely can implement a more complex logic, i.e.

mount | grep -P ' /mnt/sd .*rw'
if [ $? ]
then
mount -o remount,rw /mnt/sd
fi

But it seems the change won't make things better. We may still fail to
mount the partition as RW, which has exactly same possibility as current
solution.

What do you think?

—
Reply to this email directly or view it on GitHub
#1845 (comment).

Frenzie, there must be something I have not described clearly.
By default, the SD card is mounted as RO, i.e. koreader can access it, read without trouble. And the history information is also updated, since we write them in both history folder and sdr folder. Even sdr folder is not writable, there is no error message at all. This is a critical issue to block us from migrating history to sdr folder. So this change does not want to mount the SD card, but remount it as RW. So if mount command failed, but indeed SD card is presenting, things won't become worse. But if mount command succeeded, extra check won't make things better (users may not have sd card installed on their kobos), in such cases, 'check and do nothing' won't be better than 'try to do it, but fail to do so'.
Do you prefer to have some log about mount result, or do not start reader.lua if remount failed?

KenMaltby, are you sure it may fail on your device? Actually this command is exactly the same as what's in KSM. P.S. I am also using a Kobo AuraHD.

Tshering's scripts to remount and check the status of the external SD include feedback to the user; including:

answer=$(mount -v -o remount,rw /mnt/sd 2>&1)
case $? in
0 ) echo "remounted as read-write";;

• ) echo "$answer";;
  esac

I don't know why I had to insure the trailing "/" was there, to get it to work in recent implementations, but that is my experience.

Look at the platform/kindle/koreader.sh script if you want an example of cheap "is something already mounted?" check, where we use a bunch of them when bind-mounting custom font paths.

@Hzj-jie

But if mount command succeeded, extra check won't make things better (users may not have sd card installed on their kobos), in such cases, 'check and do nothing' won't be better than 'try to do it, but fail to do so'.

I think it would make for better debugging by not spamming error messages about an SD card that isn't there in the first place.

Do you prefer to have some log about mount result, or do not start reader.lua if remount failed?

As a user I'd rather have it working imperfectly than being faced with a draconian denial of service.

OK, based on the discussion above, I think it can be summarized as below.

• no sd card: no /mnt/sd/ found in /proc/mounts, simply ignore and do nothing
• has sd card: found /mnt/sd in /proc/mounts
  • /mnt/sd listed as RW in /proc/mounts. we are good, proceed and do nothing
  • /mnt/sd listed as RO in /proc/mounts. Try remount as RW. In case of failure, log the error, but keep bootstrapping the reader.

Let me know if I missed anything ;)

@Hzj-jie is right in that simply doing mount -o remount,rw /mnt/sd || true without all the extra condition checks will have the same result. i.e. it will only work if you have a properly formatted SD card.

That said, I too prefer we put in some extra checking logic here for couple reasons:

• to avoid unnecessary system calls. If the mount system call can be avoided with a read, why not do a grep first ;P Yes, it won't change the end result, but it's less expensive and it feels cleaner.
• it's only couple extra lines of shell commands, so trivia to do.
• keep the error handling style consistent with what @NiLuJe did in the kindle script
• it's harmless and won't break anything

Lastly, thank you @Hzj-jie for working on this legacy setting cleanup effort :)

While I have been using a remount of the external uSD card, without checks for a long time now and have had no problems related to that, it will be nice not to have to add it with each version update. I think the checks are needed, as you can never tell what the average user may encounter thru mischance or accident.

KenMaltby, no, no matter existing solution (with check) or older solution (without check), end users won't be impacted if sd card is not mountable or not existing.

I have updated this change with a check. Please kindly let me know if we still prefer '|| true' for remount command.

If you need battle-tested 'is something mounted ro?' and 'remount that rw!' checks, we have some in MRPI. It might be a tad overkill, but I'm not as familiar with how the various Kobo might mess this up to judge ;).

I can just say that something as simple as that bit us in the ass a couple of times until it was reimplemented like this on the Kindle side ;).

For the Kobos there is also the use of "Launchers", like KSM, that might be used to change the status of the external drive and of course there is Nickel and its impact on things. Also "syncthing", now.

http://www.mobileread.com/forums/showpost.php?p=3266630&postcount=122
http://www.mobileread.com/forums/showpost.php?p=3266654&postcount=123

Yes, I have a kobo aura hd myself. The machine always starts with sd card mounted as ro, so I prepared this change to make koreader sdr config work for sd card.

Thanks @NiLuJe for the pointer to MRPI, it does looks more robust. We should definitely do the same for kroeader too :)