1902020 - Single file archive base64 content length isn't being properly calculated in some cases

Status: RESOLVED FIXED

(Firefox :: Profile Backup, defect, P3)

Description

[Mike Conley (:mconley) (:⚙️)]

After bug 1897498 landed, I tried to export the simple single-file archive with my main profile. This worked properly, but when attempting to extract the ZIP file, the extraction routine choked on the base64 bytes being pulled in. When I debugged this, it looks like what's going on is that the base64 string being decoded is being truncated a bit - and this is likely because we computed an incorrect content length for the base64 bytes.

This bug is for getting that fixed and a proper test written for this case.

Comment 1

[Mike Conley (:mconley) (:⚙️)]

I found the bug. We're erroneously assuming that the number of base64 encoded bytes for some bytes X is equal to the sum of the base64 encoded bytes for some bytes X / N, where N is some chunk size.

When bytes are divided like this into chunks, if the chunk is not a multiple of 6, base64 encoding will add some padding. We need to take that individual chunk padding into account.

Comment 2

[Mike Conley (:mconley) (:⚙️)]

Attachment: Bug 1902020 - Fix a math error in calculating binary blob content length when it divides evenly by chunk size. r?#backup-reviewers!

Comment 3

[Pulsebot]

Pushed by [e-mail]:
https://hg.mozilla.org/integration/autoland/rev/45f7948e5eeb
Fix a math error in calculating binary blob content length when it divides evenly by chunk size. r=backup-reviewers,fchasen

Comment 4

[Aron Cseh]

https://hg.mozilla.org/mozilla-central/rev/45f7948e5eeb