I haven't tested on Windows.
Since alpha4 version, uploading an image seems successful, but no thumbnail nor image is generated, triggering a "This value should be of the correct primitive type" error when saving node.
Tested with png and gm.
| Comment | File | Size | Author |
|---|---|---|---|
| #19 | 2833315-19.patch | 24.74 KB | mondrake |
Comments
Comment #2
simgui8 commentedReverting 1st line in the parseFileViaIdentify() function to the alpha3 version fix the issue (at least with gm on linux).
Comment #3
simgui8 commentedPrevious patch is inverted...
Here is the right patch.
Comment #4
mondrake@guillermo-simeon thank you for your report.
1. Is this happening only with GraphicsMagick? What is your version? With GraphicsMagick 1.3.20 I cannot reproduce that.
2. Are you getting any error logged in the dblog?
3. Can you switch on the debugging option (http://cgit.drupalcode.org/imagemagick/tree/README.txt?id=deeba2c#n148) and report back commands and output?
Comment #5
mondrakeSo I checked and indeed it looks like the '%[]' syntax in the -format argument is not supported by GraphicsMagick.
The fix from @guillermo-simeon in #3 can temporarily work.
However:
1) this exposes the fact that we are missing tests for GraphicsMagick integration.
2) if we have both IM and GM installed on the server, then 'identify' may be run against IM even if the toolkit is set to use GM. This is because 'identify -format' is executed, whereas we should execute 'gm identify -format' when GM is selected.
Comment #6
mondrakeBroader patch with separate -format argument per graphics package and execution of 'gm identify -format' if GM is selected.
Add tests that may pass on DrupalCI (as the actual execution of binaries is skipped given that neither IM nor IM are installed on bots), but locally exposes some failures on GM that need to be addressed.
Comment #7
mondrakeComment #8
mondrakeComment #10
mondrakeMore tests to do.
Comment #11
mondrakeSome progress, I am not yet happy though for interaction between checkPath/imagemagickExec etc.
While running on GM the same tests that pass with IM, we have 3 points of difference that I could not work out yet:
For these I just created test bypasses, will file separate follow-up issues.
Comment #12
simgui8 commentedThanks @mondrake
The patch from #11 works fine.
Comment #13
mondrakeThanks for feedback @guillermo-simeon :)
The patch here takes the ::runOsShell method from the 8.x-2.x branch, which is now run from ::checkPath and resolves my unhappiness in #11.
Tests pass locally with both IM and GM. Adding tests for GM is a good addition to prevent future regressions.
If this is confirmed working please RTBC.
Comment #14
simgui8 commentedThe patch from #13 works for me.
Thanks @mondrake
Comment #15
mondrakeThank you. One more minor cleanup of the toolkit code and tuning of the tests.
Comment #16
simgui8 commentedPatch #15 works form me.
Nice work @mandrake
Comment #18
mondrakeOK, committed #15 to 8.x-1.x
Needs porting to 8.x-2.x
Comment #19
mondrakeComment #20
mondrakeComment #22
mondrakeCommitted #19 to 8.x-2.x
Comment #23
mondrakeReleased 8.x-1.0-alpha5.