fix for acls with usernames that have " characters in them.
This patch seems to address all the problems with usernames with double
quotes in them. The only thing I'm worried about is what happens when you
run pg_dump now against old versions of postgres that don't use the new
quoting rules...
I've done lots of dumping and loading with usernames like [\,"- _ *<> '\,\
\" \' = \" -%,*&^% and I haven't managed to break it yet. However, whomever
commits this patch should check my logic carefully, and especially check
that I have allowed properly for malloc() space for the extra characters and
stuff.
Cheers,
Chris
Attachments:
acl.txttext/plain; name=acl.txtDownload+32-23
"Christopher Kings-Lynne" <chriskl@familyhealth.com.au> writes:
This patch seems to address all the problems with usernames with double
quotes in them.
Applied, thanks.
The only thing I'm worried about is what happens when you
run pg_dump now against old versions of postgres that don't use the new
quoting rules...
AFAICS the patch would only change behavior if a username in an ACL list
contains adjacent double quotes, which seems pretty unlikely considering
the other problems a DBA would find with double quotes in user names in
older versions. So I'm not too worried.
regards, tom lane
AFAICS the patch would only change behavior if a username in an ACL list
contains adjacent double quotes, which seems pretty unlikely considering
the other problems a DBA would find with double quotes in user names in
older versions. So I'm not too worried.
Is there a place for regression tests for utils/adt? Is there some docs
somewhere that need to be updated?
Chris
Christopher Kings-Lynne <chriskl@familyhealth.com.au> writes:
Is there a place for regression tests for utils/adt? Is there some docs
somewhere that need to be updated?
We don't really have any regression tests for pg_dump, which is not a
good situation. But I wouldn't worry about this patch in particular.
As for docs, I think now it's working the way one would expect ...
if you can find anyplace that describes the old behavior, it would
need to be changed, but I don't recall that there is.
regards, tom lane