gerrit eats non-English characters in review messages

[Tom Fifield]

Hi,

It seems like gerrit doesn't play well with non-English characters - reducing them to question marks after the posting of review messages.

Steps to reproduce:
0) Set encoding to UTF-8 in browser (should already be set by content="text/html; charset=UTF-8" in header) and install appropriate fonts
1) Open a patch in review in gerrit
2) Click review
3) Type non-English characters in the "cover message" box and click submit

What should happen:
English and non-English characters are displayed as they were typed into the "cover message box"

What actually happens:
Question marks are displayed in place of non-English characters

See https://review.openstack.org/#/c/25313/1 for an example:

???????????? ?cinder_huawei_conf.xml??????‘<'?????‘>' ???‘>')???

[Jeremy Stanley (fungi)]

I notice that glyphs having an analog in ISO/IEC 8859-1 display fine, but (for example) Hiragana does not. This is almost certainly due to Gerrit reencoding between UTF-8 and its underlying MySQL table encoding. Thus I've marked this as a duplicate of our bug to reencode freeform text fields in Gerrit's database.