Details
-
Type:
Bug
-
Status: Closed
-
Priority:
Major
-
Resolution: Fixed
-
Affects Version/s: 4.2.2
-
Fix Version/s: 5.0.0 alpha1, 5.0
-
Component/s: Core/Parsing
-
Labels:None
-
Environment:any
-
ICEsoft Forum Reference:
Description
Tasks for this bug.
- figure out why the fallback code isn't executing
- consider an parser enhancement, instead of parsing the whole file into memory on failure, we should consider updating the xref table with new byte offset found by parsing file, but not keeping the objects in memory.
Activity
- All
- Comments
- History
- Activity
- Remote Attachments
- Subversion
This represent quite a big block of work and would be really nice feature. Moving to 5.0.
Changes made for PDF-410, have correct the load issues with graph.pdf. The document Linerized.pdf is still a problem.
Quite a bit of work has already been done to improve how we handle file that have incorrect xrefs. Under the new model if the catalog can not be found we fall back to linear read that also keeps track of the actual object offsets and rebuilds the xref table. Once the xref table is rebuild it's possible to re-fetch the objects lazily as the Page references are garbage collected.
Of the attached documents the graph.pdf no loads correctly but there are some logged exceptions that need to be cleaned up. The second pdf still does not load so I'll used it to review our fall back logic.
Updated catalog initialization to try an force any objects offset errors to show up while we still have a change to fall back to a linear traversal and subsequent reindex of the cross reference table.
Xref problems.