Change libhunspell's file access code that loads dictionaries to use function pointers
Categories
(Core :: Spelling checker, enhancement, P3)
Tracking
()
Tracking | Status | |
---|---|---|
firefox81 | --- | fixed |
People
(Reporter: shravanrn, Assigned: shravanrn)
References
Details
Attachments
(1 file, 1 obsolete file)
Hunspell is the spelling checking library used in Firefox. Hunspell loads dictionaries from the file system. Currently this file system operations are applied via a in-tree patch to hunspell which makes direct calls to NS_* apis to open and read files from URIs.
Hunspell is to be sandboxed using RLBox. See Bug 1653659. In preparation for this, all NS_* calls for file access should be exposed via callbacks.
Assignee | ||
Comment 1•5 years ago
|
||
Assignee | ||
Comment 2•5 years ago
|
||
Updated•5 years ago
|
Comment 4•5 years ago
|
||
Only skimmed the patch, but doesn't this [1] blindly trust the filename from the sandbox? Wouldn't it be better for the filename to be managed entirely by the code outside the sandbox?
[1] https://hg.mozilla.org/integration/autoland/rev/28fe4f755717#l4.82
Comment 5•5 years ago
|
||
bugherder |
Assignee | ||
Comment 6•5 years ago
|
||
@bholley : yup! That will be addressed in a separate patch. This patch simply makes some refactoring changes that makes the library easier to RLBox (namely separating the file access bits from the rest of hunspell). In the subsequent patches, (which are 70% complete) when we use the RLBox api, we will automatically be required to use file paths only from a trusted source.
Description
•