PagedTiffHandler uses ImageMagick identify to get some information about the file, such as the widths and heights of each of the pages. It appears to be extremely memory-inefficient, probably decompressing the whole image. This will cause images to fail to be rescaled, or to fail to upload, once PagedTiffHandler is deployed.
Please consider tiffinfo from libtiff as an alternative.
Version: unspecified
Severity: normal
| Status | Subtype | Assigned | Task | ||
|---|---|---|---|---|---|
| Resolved | None | T25258 Enable PagedTiffHandler on all wikis, to allow display of TIFF files | |||
| Resolved | Mglaser | T26821 ImageMagick identify uses a lot of memory |
I understand the issue, and I'll llook into libtiff for this. However, using identify has one very important advantage: it will fail just in the cases where convert/mogrify will fail. That is, it's a pretty good way to determine which tiff files can actually be rendered, and which can't.
Actually it won't. We use a different shell memory limit on the image scalers compared to the main cluster.
I wasn't talking about failure because of memory limits, but because of unsupported variants of tiff - codecs, etc. there's quite a few of those, check out http://www.libtiff.org/images.html
marking as fixed, since libtiff/tiffinfo is now used live. Issues with tiffinfo are being tracked by bug 24984.