It rather sounds like this was a bug in 10g that has been fixed in 11g.
You see stuff like that occasionally.
A personal favorite of that was a VIEW that compiled successfully in 9i,
and had been used for years. On testing a migration to 10g, we found the
view would not compile in 10g because one of the columns referenced was
ambiguous.
Checking the definition, we found that the code was wrong -- the column in
question was definitely ambiguous.
I explained this to my customer -- that their code was in error, and needed
to be corrected. The view should never have been usable on 9i, and there
was even a chance they were getting wrong (or unintended) results.
The customer demanded that we instead make the view compile as-is on 10g
-- in effect they wanted us to ask Oracle to forward-port a bug ! The
upgrade was eventually cancelled, and the database runs on 9i to this day.
On Fri, Apr 17, 2015 at 10:18 AM, Patrice sur GMail <
patrice.boivin@xxxxxxxxx> wrote:
I noticed that in 10g, one could recompile a package body with invalid
database link references, but that 11g is more thorough: If a reference no
longer works it refuses to compile the package body.
-- Patrice
My profiles: [image: Facebook]
<http://www.facebook.com/home.php?#!/profile.php?id=100000206805521>[image:
LinkedIn] <http://ca.linkedin.com/pub/patrice-boivin/a/933/5a9>[image:
Twitter] <http://www.twitter.com/PatriceBoivin>
<http://www.twitter.com/PatriceBoivin>
Signature powered by WiseStamp <http://www.wisestamp.com/email-install>