Do cu specs #113
Do cu specs #113
Conversation
gnat2goto/driver/tree_walk.adb
Outdated
| Unit_Name = "system%s" | ||
| then | ||
| null; | ||
| -- At the moment Standard or System are not processed - to be done |
There was a problem hiding this comment.
Maybe some Put_Line warning about this?
In the message not sure if 'package' or 'unit' might be more appropriate
Put_Line (Standard_Error,
"At the moment Standard or System library units are not processed. ");
There was a problem hiding this comment.
I take it back: not processing standard/system libraries is not a problem unless you rely on something in them (and then you would get the appropriate error message).
There was a problem hiding this comment.
@xbauch Nitpick:
Okay. The "if then else" construct begs the question.
To prevent another reader wondering why the null branch of the "if then else" rather than just just "if not" maybe add a clarifying line?
I'll assume the if else then construct is in case in the future we want to consider doing something with standard/system libraries.
| case Nkind (N) is | ||
| when N_Subprogram_Body => | ||
| if Acts_As_Spec (N) then | ||
| -- The unit is a withed library unit which subprogram body |
There was a problem hiding this comment.
suggest add summary line at top (similar to this)
-- withed library unit that has no separate declaration
(this gives the reader what is common among the scenario discussed below)
xbauch
force-pushed
the
do_cu_specs
branch
2 times, most recently
from
6184ba2 to
a8b77a2
Compare
|
@sonodtt and @hannes-steffenhagen-diffblue I've addressed the comments made here and on xbauch#2. I have added commit messages summarising their content. |
gnat2goto/driver/tree_walk.adb
Outdated
| else | ||
|
|
||
| -- Withed library unit that has no separate declaration |
There was a problem hiding this comment.
I do not understand this comment. These withed library units do have library-level declarations. Any user defined library-level unit must have an explicit declaration whether it is a package, subprogram declaration or just a subprogram body. This is what I tried to explain in the comments under each case.
A more general comment preceding the case statement might be
"-- Handle all other withed library unit declarations"
| @@ -204,6 +207,12 @@ package body Tree_Walk is | |||
| function Do_Op_Not (N : Node_Id) return Irep | |||
| with Pre => Nkind (N) in N_Op; | |||
|
|
|||
| procedure Do_Package_Declaration (N : Node_Id) | |||
There was a problem hiding this comment.
I seemed to have introduced a number of procedures starting with Do_* which is not in accordance with the naming convention according to xbauch#2 (comment).
I don't think its worth changing now because it should mean re-ordering the declarations again to keep them in alphabetical order. If it is ok with everyone merge with these names now and I will sort out the naming conventions when I do the next PR.
There was a problem hiding this comment.
@tjj2017 I have a small number of queries that are also just minor nitpicks / request for clarification (often just due to my poor knowledge of ADA) that I think should not get in the way of merging what is a great body of work.
I'll ask about these separately - if additional comments are helpful then they can be in a separate minor PR.
xbauch
force-pushed
the
do_cu_specs
branch
2 times, most recently
from
89c994a to
34cd22b
Compare
| @@ -226,6 +226,17 @@ package body Driver is | |||
| end; | |||
|
|
|||
| if not Add_Start then | |||
There was a problem hiding this comment.
Default mode "don't add comments" to this thing I had to figure out what it is doing by looking at the the out parameter in another function.
Status: "on".
:)
There was a problem hiding this comment.
Fixed, and noted for the next time.
Introduces Register_Subprogram_Specification procedure that creates a new symbol table entry with the subprogram name.
Process_Declaration(s) mirror the behaviour of Process_Statement(s), i.e. a case analysis on the kind of declaration stored in the node being processed. This commit also extends the cases in Process_Statement.
Introduces: 1) Do_Package_Declaration 2) Do_Package_Specification -- processes private and visible declarations as a new code block 3) Do_Withed_Unit_Spec -- case analysis for registering program specification/processing subprogram declarations/package declarations 4) Do_Withed_Units_Specs
Also includes cprover_start even when there is no function to call in a CU. Plus one more failing test (package function bodies are not processed).
Mostly Trevor's work, the last commit addresses some issues from xbauch#2 and adds a test for packages.