Testing suggestion

[lurch]

Maybe some of the issues we've run into with drivelist, are also things that should be actively tested with mountutils?

[jviotti]

I agree. Lets compile a list of things to test here, and we can the move it to a markdown documentation in the repo, as a manual testing guide:

Drive mount points

• Drives whose mount points contain spaces
• Drives whose mount points contain quotes (single and double)
• Drives whose mount points contain symbols such as #, and &
• Unmount a drive with multiple mount points

Drive types

• Unmount an SD Card using an internal reader
• Unmount an SD Card using an external reader
• Unmount a removable USB drive
• Unmount a USB HDD

Edge cases

• Unmount an empty internal reader device on Windows
• Unmount a drive that doesn't exist
• Unmount a drive without the necessary permissions
• Unmount an already unmounted drive
• Unmount a drive which has open handles on it
• Unmount a drive by running mountutils from inside the same drive

[lurch]

@jviotti Do you remember what the issue you had was, where you were getting different behaviour when the SD-reader was plugged in via a USB3 hub, than when it was plugged in directly?

Any manual testing guide should also include the things you've already been testing this module against, e.g. unmounting non-existent drives / devices, unmounting already-unmounted devices, unmounting drives with multiple-mounted partitions, etc. etc.
There's (probably) going to be so many edge cases, that IMHO the more different things we can add to a list of manual tests, the better. When you're not running through a checklist, it's easy to remember to test one thing on one OS, but then forget to test that same thing on a different OS.

I wonder if we could do any (basic) automated Linux testing by maybe scripting Qemu?

[jviotti]

@jviotti Do you remember what the issue you had was, where you were getting different behaviour when the SD-reader was plugged in via a USB3 hub, than when it was plugged in directly?

I'm afraid I don't :( Do you remember if this was mentioned in a GH issue, or in the Gitter channel?

Any manual testing guide should also include the things you've already been testing this module against, e.g. unmounting non-existent drives / devices, unmounting already-unmounted devices, unmounting drives with multiple-mounted partitions, etc. etc.
There's (probably) going to be so many edge cases, that IMHO the more different things we can add to a list of manual tests, the better. When you're not running through a checklist, it's easy to remember to test one thing on one OS, but then forget to test that same thing on a different OS.

I agree, let me update the list above with what I've been testing so far.

I wonder if we could do any (basic) automated Linux testing by maybe scripting Qemu?

Sounds interesting. What would we automatically mount in the VM though? Can we somehow fake a drive?

[lurch]

I'm afraid I don't :( Do you remember if this was mentioned in a GH issue, or in the Gitter channel?

Don't remember. I tried searching both of those and can't find anything relevant :(

I wonder if we could do any (basic) automated Linux testing by maybe scripting Qemu?

Sounds interesting. What would we automatically mount in the VM though? Can we somehow fake a drive?

Yup, in theory Qemu should allow you to fake anything ;-) A quick search finds https://wiki.ubuntu.com/QemuDiskHotplug

[lurch]

Should we also test unmounting drives which have open file-handles? I guess it should report "drive is busy" or something on all OSes?

[jviotti]

Yup, in theory Qemu should allow you to fake anything ;-) A quick search finds https://wiki.ubuntu.com/QemuDiskHotplug

Sounds exciting. I'm happy to experiment with it.

Should we also test unmounting drives which have open file-handles? I guess it should report "drive is busy" or something on all OSes?

Good catch. In fact, all OSes will try to close the open handles before unmounting, and only fail if they can't do so.

[lurch]

I guess an easy way to test the un-closable handles behaviour would be to try running mountutils from the drive we're actually trying to unmount! ;-)

[jviotti]

How evil! :) I'll add that to the list as well.