Re: tar-related code in PostgreSQL

Robert Haas <robertmhaas@gmail.com> writes:
> We have similar code in many places -- because evidently nobody
> thought it would be a good idea to have all the logic for reading and
> writing tarfiles in a centralized location rather than having many
> copies of it -- and typically it's written to pad the block out to a
> multiple of 512 bytes. But here, the file is 0 bytes long, and then we
> add 511 zero bytes. This results in a tarfile whose length is not a
> multiple of the TAR block size:

Bleah.  Whether or not the nearest copy of tar happens to spit up on
that, it's a clear violation of the POSIX standard for tar files.
I'd vote for back-patching your 0001.

I'd lean mildly to holding 0002 until after we branch.  It probably
won't break anything, but it probably won't fix anything either.

			regards, tom lane