libencode-perl 2.82-1: MIME encoding broken on 16.04

[Mihai Moldovan]

lsb_release -rd:

Description: Ubuntu 16.04.2 LTS
Release: 16.04

I'm using perl for encoding the From: header value as part of a git hook with the author's email address.

This worked fine on 12.04, but fails badly on 16.04.

Example:

echo "Foooooo Baaaaaar <email address hidden>" | perl -CI -MEncode -e 'print Encode::encode("MIME-Q", <STDIN>)'
=> =?UTF-8?Q?Foooooo=20Baaaaaar=20=3Cfoo?= =?UTF-8?Q?=40bar=2Ebaz=3E?=

The string split and additional added whitespace means that MTA's see an address (after decoding) of "Foooooo Baaaaaar <foo @bar.baz>" and thus break in horrible ways.

If characters are removed (preferably from the "real name" part), the split is shifted or at best averted - but typically one cannot limit author addresses to a very limited fixed size.

The same command executed on OS X 10.9 with MacPort's perl 5.24.1 and Encode 2.88 is working fine:

echo "Foooooo Baaaaaar <email address hidden>" | perl -CI -MEncode -e 'print Encode::encode("MIME-Q", <STDIN>)'
=?UTF-8?Q?Foooooo_Baaaaaar_=3Cfoo=40bar=2Ebaz=3E=0A?=

This is generally what I'd expect as well - no string splitting in-between. Given that we try to avoid patching software whenever possible in MacPorts, this is a "close-to-vanilla" test, although obviously the Encode module at version 2.88 is different than what libencode-perl ships - at version 2.82_1.

The same split up also happens on Debian unstable, which ships the Encode module version 2.80_1 as part of its libperl5.24 package.
If libencode-perl (2.88-1) is installed on Debian unstable - overriding the core module version, this bug does *not* occur.

On Ubuntu 16.04, the Encode module as a Perl core module (i.e., without libencode-perl being installed), does not have any capabilities of MIME conversion, so testing without libencode-perl is impossible for the scope of this report.

Other Ubuntu versions might also be affected. I have only tested 16.04.

[Florian Schlichting (fschlich)]

As you note this is fixed as of Encode 2.88 (but see the discussion in https://github.com/dankogai/p5-encode/pull/68), so I'm marking this bug as "fix released"