"Runaway" mmd processes

[fletcher]

Several users have reported that they have experienced several instances of multimarkdown processes being spawned that start to use 100% of their CPU.

1. This is in all likelihood not really a multimarkdown issue per se, but rather an issue involving another application that is calling mmd. multimarkdown itself is a command-line program that has no way of calling itself. It doesn't spin off multiple instances of itself.

2. I don't know why mmd, once spawned, is not ending, but without more information about the programming calling it, and what sort of input it is getting, I can't tell if this is truly a bug or bad input.

3. If you experience this, you need to pay attention to other programs that can cause the multimarkdown command to run --- e.g. Quicklook, Marked, and others. You can try removing the Quicklook generator - that has spawned a runaway process for me once or twice when I accidentally tried to preview a really long text file

4. If this happens, you can use the Activity Monitor application to kill the runaway processes.

5. If anyone can reliably reproduce this issue to help me track it down, I will be quite grateful. But for now, it appears that the only way to prevent it is to disable the other applications until something more specific can be found.

But to be clear - this is not a problem with multimarkdown replicating itself on its own and running amok. Something has to be launching the multimarkdown command, and that's where you need to look.

[fletcher]

Another clue - was looking at icons in the Finder and realized that if you have the QuickLook generator installed, it will use the generator to create a preview for the icon.

This means that the quicklook generator could be called without you necessarily realizing it.

[ttscoff]

I found the Quicklook generator to be the issue on my system. At first I thought it was Byword, but appeared to happen only when the mds process spun up for a refresh.

[fletcher]

Perhaps now that I've been spending so much time in "Cocoa-Land", I'll take another look at the QuickLook code and try to make it more robust.

[fletcher]

I created a brand new version of the quicklook plugin to see if I can track this down. New version includes built-in mmd (3.2 for i386). Previews both plain text (with the .md extension) and OPML files. It also has minimal console logging to potentially help figure out the problem.

Check it out if you want to test, and let's see if we can get this problem solved.

http://d.pr/otTP

F-

[ttscoff]

Running for 12 hours now with no issues, will keep you posted.

[fletcher]

How often did you have issues before? My problem is that the runaway processes were very infrequent, which makes it hard for me to know they are not happening as opposed to not happening yet.

But so far, it's still running well for me also.

[ttscoff]

They would spin up every couple of hours, I'd kill them, they'd come back. Not seeing that at all this time around.

[fletcher]

New version of QuickLook has been pushed to github. Old version removed. I am going to close this issue here.

If it recurs, this can be reopened in the new MMD-QuickLook repo. But hopefully that won't be necessary....