Created attachment 49486 [details]
In the 3.4.2 rc2 the import of some components of graphics from MS Power point got broken.
Gustavo_PE.ppt - original file
Gustavo_PE_342RC2.pdf - pdf export from original file in 3.4.2 RC2
Gustavo_PE_341.pdf - pdf export from original file in 3.4.1
Created attachment 49487 [details]
PDF export with 3.4.2 RC2
Created attachment 49488 [details]
PDF export with 3.4.1
It looks like there is another problems with the attached document. The first chart seems to be sized incorrectly (the line graph). So when you export to PDF directly, it does not show up (because you only have the white portion on the top left of the graph). Re-sizing seems to cause problems as well, but I do not know why. The afterimage of the previous graph remains. Well, after adjusting the size of the image, the pdf export shows the change as well, I presume that the export is ok. The bar graph below shows up when I export fine.
So there is a problem with graphic components. Do you encounter a similar problem with your file?
Also, it looks like you were going to attach a third file but you haven't...
Created attachment 49490 [details]
What I see after readjusting graph and exporting to PDF
Sorry I did not see your third export until after I posted the first comment. It does look like the problem is pretty big because the numbers on the line graph are lost. I am using LibreOffice 3.4 340m1(Build:103)on OpenSuse Linux.
I confirm the behaviour as described by Branislav on Ubuntu.
The problems that you have described (resining..) are also occurring. The funny thing is that in 3.4.1 all looks fine.
I have similar problems also with other ppt files. This one was just an example.
let me get a build & look into this one.
Ah, this looks like a serious regression and 3.4.2 release blocker.
Created attachment 49532 [details]
apparently this is emf+ comment parsing related. Looking through the emf+ records, it looks like the assumption made that the minimum valid record length is 12 + non-0 length payload isnt' right, looks like that the minimum valid record length is just 12, i.e. no need for a payload.
Created attachment 49533 [details]
real patch, not pseudo-code this time
(In reply to comment #9)
> Created an attachment (id=49532) [details]
> probable fix
> apparently this is emf+ comment parsing related. Looking through the emf+
> records, it looks like the assumption made that the minimum valid record length
> is 12 + non-0 length payload isnt' right, looks like that the minimum valid
> record length is just 12, i.e. no need for a payload.
Yep, to the point. grabbed for master & -3-4, need two more reviews for 3-4-2 now.
Approved and pushed into 3-4-2 for rc3.