Pragmas in declarations (volatile) [depends-on: #220] #215
Conversation
| Data := 0; | ||
| end if; | ||
|
|
||
| pragma Assert (Data=42); |
There was a problem hiding this comment.
What was the behaviour of symex prior to the previous changes? Would this particular test have failed anyway?
There was a problem hiding this comment.
No, it would look exactly the same way. Only the symbol table would contain the volatile flag and gnat2goto would complain about not supporting pragma volatile.
There was a problem hiding this comment.
Sorry to be a pain but please can we have some more test cases for the other issues that this seems to have resolved. I think a number of them have open tickets and it is not clear to me why the changes in this PR affect things like N_Attribute_Reference or N_Integer_Literal.
The handling of volatile seems fine for now.
xbauch
force-pushed
the
feature/pragma-declaration-volatile
branch
from
645070d to
1ef713c
Compare
xbauch
changed the title
|
@martin-cs Those changes come from the first commit, where we store the identifier being defined in the symbol table. I created a stand-alone PR for that change: #220. |
via a separate method that looks-up the symbol and updates it.
to check that the symex does not choke (the symbols have the volatile flag set in the symbol table).
xbauch
force-pushed
the
feature/pragma-declaration-volatile
branch
from
1ef713c to
789d901
Compare
xbauch
changed the title
|
I'll rebase this branch as a new PR and merge. I'll post a new comment here when it's done. |
|
The rebased version is in #246. |
|
Merged in #246. |
Handle volatile pragmas by updating the symbol in symbol table. Turned out the symbol wasn't there yet. Extended the
Do_Defining_Identifierto insert the symbol.The first commit is a squash of the prerequisite PR #220 -- no need to review it here.