> On 12 Aug 2018, at 11:01, Andres Freund <[email protected]> wrote: > > On August 12, 2018 12:17:59 AM GMT+02:00, Daniel Gustafsson <[email protected]> > wrote: >>> On 6 Aug 2018, at 09:47, Heikki Linnakangas <[email protected]> wrote: >>> >>> Has there been any consideration to encodings? >> >> Thats a good point, no =/ >> >>> What happens if the message contains non-ASCII characters, and the >> sending backend is connected to database that uses a different encoding >> than the backend being signaled? >> >> In the current state of the patch, instead of the message you get: >> >> FATAL: character with byte sequence 0xe3 0x82 0xbd in encoding "UTF8" >> has >> no equivalent in encoding “ISO_8859_5" >> >> Thats clearly not good enough, but I’m not entirely sure what would be >> the best >> way forward. Restrict messages to only be in SQL_ASCII? Store the >> encoding of >> the message and check the encoding of the receiving backend before >> issuing it >> for a valid conversion, falling back to no message in case there is >> none? >> Neither seems terribly appealing, do you have any better suggestions? > > Restricting to ASCII seems reasonable.
It’s quite restrictive, but it’s the safe option. I’ve hacked this into the updated patch, but kept the backend_feedback() function using pg_mbstrlen() at least for now since it seems the safe option should this be relaxed at some point. Also added a small test by copying text from a ja.po file in the tree. > But note that sqlascii isn't that (it's essentially just arbitrary null > terminated data). Easier to relax later. Yeah, my fingers and brain were not in sync during typing, I meant to say ASCII there. I blame a lack of coffee. cheers ./daniel
0001-Refactor-backend-signalling-code-v14.patch
Description: Binary data
0002-Support-optional-message-in-backend-cancel-terminate-v14.patch
Description: Binary data
