CVS difference for ai12s/ai12-0247-1.txt

--- ai12s/ai12-0247-1.txt	2018/01/19 05:05:51	1.1
+++ ai12s/ai12-0247-1.txt	2018/01/27 04:54:51	1.2
@@ -11,8 +11,9 @@
 !subject Potentially Blocking goes too far for Detect_Blocking
 !summary
 
-During a protected action, a call on a subprogram that contains a potentially
-blocking operation, may, but does not have to, raise Program_Error.
+During a protected action, a call on a subprogram that contains a
+potentially blocking operation, may, but does not have to, raise
+Program_Error.
  
 !question
 
@@ -374,6 +375,90 @@
 explains why this exception is happening. Hopefully, everyone will like it.
 
 As always, comments welcome.
+
+****************************************************************
+
+From: Jean-Pierre Rosen
+Sent: Friday, January 19, 2018  3:17 AM
+
+> !ACATS test
+> 
+> Bounded errors are not usefully testable.
+
+If Detect_Blocking applies, then it's no more a bounded error, and a test is
+necessary. Or is the test already there (sorry, I didn't care to check) ?
+
+****************************************************************
+
+From: Tucker Taft
+Sent: Friday, January 19, 2018  7:33 AM
+
+Seems like a nice solution.
+
+****************************************************************
+
+From: Bob Duff
+Sent: Friday, January 19, 2018  7:50 AM
+
+I agree.
+
+****************************************************************
+
+From: Tullio Vardanega
+Sent: Friday, January 19, 2018  9:04 AM
+
+Neat, I like it.
+Thanks Randy.
+
+****************************************************************
+
+From: Randy Brukardt
+Sent: Tuesday, January 23, 2018  12:25 AM
+
+> > Bounded errors are not usefully testable.
+> If Detect_Blocking applies, then it's no more a bounded error, and a 
+> test is necessary. Or is the test already there (sorry, I didn't care 
+> to check) ?
+
+Any such test clearly belongs to pragma Detect_Blocking, as all this AI does
+is reduce the cases where Detect_Blocking is required to do something. So the
+most that might need to be done is modifying an existing test. I don't
+generally mention that in these clauses, as it applies to any Binding
+Interpretation AI (some existing test might need to be changed).
+
+Now, is your question whether there are any Detect_Blocking tests? The answer
+is probably not; all of the real-time tests were farmed out to IRTAW which has
+to date resulted in exactly 0 ACATS tests. I've been prioritizing the core
+language (which Detect_Blocking is not) in my own test writing. Annex C is
+about as far back in the Standard as I've gone. In any case, test(s) for
+Detect_Blocking is on the Ada 2005 to-do list (with a priority 8 out of 10).
+
+P.S. If such a test showed up in my in-box, it would probably get issued ASAP.
+I'm mostly working on priority 8 issues at this point, so it is among the
+highest priority issues remaining.
+
+****************************************************************
+
+From: Randy Brukardt
+Sent: Friday, January 19, 2018  12:36 AM
+
+> If Detect_Blocking applies, then it's no more a bounded error, and a 
+> test is necessary. Or is the test already there (sorry, I didn't care 
+> to
+> check) ?
+
+My earlier answer probably answered the question you meant, but it's
+irrelevant for this AI.
+
+This AI defines a new bounded error for a specific case. That error has
+nothing to do with Detect_Blocking (that was the entire point of the change)
+and as such any testing of this AI also has nothing to do with
+Detect_Blocking. So there is only a bounded error, which is not very valuable
+to test. (This particular error could be tested, but generally ACATS tests are
+not supposed to care as to what implementation choices are made when the
+Standard gives choices. And it's unclear what incorrect results could be
+tested; guessing about incorrect results are not valuable.) Therefore, Bounded
+Errors are interesting only as things to avoid in ACATS tests.
 
 ****************************************************************