Re: pg_waldump: support decoding of WAL inside tarfile

Andrew Dunstan <andrew@dunslane.net> writes:
> I also noticed a possible bug in astreamer, where the decompressor 
> finalize functions send bbs_buffer.maxlen bytes to the next streamer 
> when flushing remaining data at end-of-stream. This seems wrong because 
> the buffer may only be partially filled with valid decompressed data. 
> Possible patch for that attached. (But I don't think it's related to 
> these failures).

Ugh.  That's surely very broken, but how did we not notice?
Do all the consumers know enough to ignore garbage trailing data?

            regards, tom lane