This is the 8.x-1.x version of the issue with the same name for 7.x. This same deadlock bug applies to both.
| Comment | File | Size | Author |
|---|---|---|---|
| #4 | deadlock-8.x-2897356-4.patch | 955 bytes | jonathan.green |
| #2 | deadlock-2897356-2.patch | 999 bytes | jonathan.green |
This is the 8.x-1.x version of the issue with the same name for 7.x. This same deadlock bug applies to both.
| Comment | File | Size | Author |
|---|---|---|---|
| #4 | deadlock-8.x-2897356-4.patch | 955 bytes | jonathan.green |
| #2 | deadlock-2897356-2.patch | 999 bytes | jonathan.green |
Comments
Comment #2
jonathan.green commentedThis is the same issue as in the 7.x branch, the code used to run imagemagick is basically the same. This is a port of the patch I made for the 7.x branch.
Original Comment:
Comment #3
mondrakeThank you @jonathan.green
We need to fix this in 8.x-2.x first. The patch does not apply there, unfortunately, as that piece of code was moved to the
ImagemagickExecManagerclass.Also, it would be great to have tests to demonstrate the failure.
Comment #4
jonathan.green commentedHere is the patch updated to 8.x-2.x, the code is again basically the same, just in a different place.
Testing this is tricky. I think it needs more testing-fu then I have. Maybe something like shipping with a bad PDF in the test folder then running a convert on it. Not sure exactly. I'm fairly convinced it will depend on the system installed versions of IM, GS and the libopenjp2, because that is what determines how much output goes to stdout.
Comment #5
jonathan.green commentedComment #6
mondrakeThanks @jonathan.green
If anyone can post here an offending PDF (or other formats like broken JPEG as I saw in the D7 version of this issue), of course assuming that it is OK to put in the git repo as a test file, it may help.
A concern I have is that the warning message is logged/presented on screen and would like to ensure a 64k+ string does not create other issues.
Comment #7
mondrakeCan anybody check if the latest patch in #2911289: Use Symfony process component instead of proc_open would solve this issue?
Comment #8
mondrakeThis issue should be overcome in the latest 8.x-2.x-dev release after the commit of #2911289: Use Symfony process component instead of proc_open. Please report back if not.