Horton

Re: Decision by Monday: PQescapeString() vs. encoding violation

Tom Lane <tgl@sss.pgh.pa.us>

View thread View on postgr.es
From: Tom Lane <tgl@sss.pgh.pa.us>
To: Andres Freund <andres@anarazel.de>
Cc: Andrew Dunstan <andrew@dunslane.net>, Noah Misch <noah@leadboat.com>, pgsql-hackers@postgresql.org, pgsql@j-davis.com
Date: 2025-02-15T18:35:10Z
Lists: pgsql-hackers 
Andres Freund <andres@anarazel.de> writes:
> It seems that nobody is arguing against the "just skip one byte" behaviour, so
> I'm inclined to push this fairly soon, even if Noah's "24 hours" haven't quite
> elapsed.  A few more cycles in the buildfarm wouldn't hurt.

Agreed.  I thought there would be more discussion, but it seems
nobody really objects to changing this.

The other thing that was discussed in the security thread was
modifying PQescapeStringInternal and PQescapeInternal to produce
no more than one complaint about invalid multibyte characters,
on the grounds that input that's just plain in some other encoding
would otherwise produce a ton of repetitive messages.  That seems
trivial enough to mechanize with a bool already_complained flag,
so I think we should incorporate that refinement while we're here.
I can write that patch if you're busy.

			regards, tom lane

Commits

  1. Make escaping functions retain trailing bytes of an invalid character.

  2. Fix handling of invalidly encoded data in escaping functions