425768 - Zipwriter corrupts zip files that previously contained extra fields

Status: RESOLVED FIXED

(Core :: General, defect)

Description

[Dave Townsend [:mossop]]

When editing a zip file created by a third party tool that wrote extra field data for any entry in the zip file, zipwriter will not correctly output the zip file leaving it unusable.

The fix for this is fairly simple, not going to be prone to cause any regression and can easily be unit tested.

Comment 1

[Dave Townsend [:mossop]]

Attachment: patch rev 1

This is some poor code that went effectively untested because the extra field is unused by zipwriter which writes pure zip files. The already in tree testcase of this didn't flag any errors because every unit test removes existing things from it. This new unit test just adds something and checks that the file isn't corrupt (it is without this patch) and that the length is correct.

The patch just properly uses the mFieldLength field to keep the length of the extrafield known.

Comment 2

[Mike Beltzner [:beltzner, not reading bugmail]]

Comment on attachment 312356 [details] [diff] [review]
patch rev 1

a=beltzner

Comment 3

[Dave Townsend [:mossop]]

Checking in modules/libjar/zipwriter/src/nsZipHeader.cpp;
/cvsroot/mozilla/modules/libjar/zipwriter/src/nsZipHeader.cpp,v  <--  nsZipHeader.cpp
new revision: 1.2; previous revision: 1.1
done
CS file: /cvsroot/mozilla/modules/libjar/zipwriter/test/unit/test_bug425768.js,v
done
Checking in modules/libjar/zipwriter/test/unit/test_bug425768.js;
/cvsroot/mozilla/modules/libjar/zipwriter/test/unit/test_bug425768.js,v  <--  test_bug425768.js
initial revision: 1.1
done