Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

__FL__ added and tested #3574

Closed
wants to merge 4 commits into from
Closed

Conversation

briangmilnes
Copy link
Contributor

Two changes: 1) FL lexeme in FStarC_Parser_LexFStar.ml
2) tests/validation-time with a Test.LexemeFL.fst that checks FL

…alidation-time with a Test.LexemeFL.fst that checks __FL__
@briangmilnes
Copy link
Contributor Author

Is this done even if it failed to post to slack? I think I'm confgured now.

@@ -473,6 +473,7 @@ match%sedlex lexbuf with
| "#print-effects-graph" -> PRAGMA_PRINT_EFFECTS_GRAPH
| "__SOURCE_FILE__" -> STRING (L.source_file lexbuf)
| "__LINE__" -> INT (string_of_int (L.current_line lexbuf), false)
| "__FL__" -> STRING ((L.source_file lexbuf) ^ "(" ^ (string_of_int (L.current_line lexbuf)) ^ ")")
Copy link
Member

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

__FL__ seems a bit obscure to me, should we maybe call this __FILELINE__? But I don't have a strong opinion, would be good if others chime in.

Copy link
Contributor Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

You have to reference it in every test so long is kind of cumbersome. For example:
let test_system n =
let r = FStar.Unix.system "echo HI" in
match r with
| WEXITED e -> if_test FL (e = 0) n None
| _ -> final_fail FL n None

Copy link
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

I agree with Guido, I aslo find FL a bit obscure.
IMO, aggregating LINE and SOURCE_FILE in this formatting is quite opinionated and should probably be user-defined.

@briangmilnes why not using a tactic for that here, e.g.:

module Hello
open FStar.Char
open FStar.Tactics

let fl
  (x: unit)
  (#[let sealed_range = range_of_term (quote x) in
     let range = unseal sealed_range in
     let result = FStar.Range.explode range in
     exact (quote result)] loc: string * int * int * int * int)
 = let file, line_start, col_start, _line_end, _col_end = loc in
   file ^ "(" ^ string_of_int line_start ^ ")"

// This is line 14
let file_line: string = fl ()
let _ = assert_norm (file_line == "<input>(15)")

Copy link
Contributor Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

This is for logging and test packages. Nothing should be left to the user. In testing the ocaml wrap of unix and a few other things, I need the file/line 139 times. Terse is good here, although I eschew it in almost all situations.

A tactic that allows a function is a great alternative. The language server in emacs takes that and works but I can't get a command line to compile it. What's the change needed for compilation? It also has the advantage that the strings are not coming back from the lexer with strange UTF8.

Copy link
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Well, verbosity or terseness is a choice, right? I tend to prefer verbosity in such cases, but that's only my preference.

Nice! Here it won't work via the command line because of <input>: that's the file name given by the F* interactive protocol used by emacs I believe.
Running that from a file whose path is /tmp/Hello.fst, you should be able to assert_norm (file_line == "/tmp/Hello.fst(15)") (instead of the one I've put above).

Copy link
Contributor Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

The issue I have is it will not compile, not the string back from fl() or FL.

F*orge: VALIDATE _build/fstar/fst/checked/Hello.fst.checked

  • Error 168 at src/Hello.fst(8,8-8,8):
    • Syntax error

1 error was reported (see above)
That's the (# line on my test of it running in OCaml.

tests/Makefile Outdated
@@ -21,6 +21,7 @@ ALL_TEST_DIRS += vale
ALL_TEST_DIRS += hacl
ALL_TEST_DIRS += simple_hello
ALL_TEST_DIRS += dune_hello
ALL_TEST_DIRS += validation-time
Copy link
Member

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

The new single test here is not related to time at all? I think you can just drop Text.LexemeFL.fst in tests/micro-benchmarks and not add any new Makefiles.

Copy link
Contributor Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

validation-time is the name of the phase right? Nothing to do with time. I didn't put it in micro-benchmarks as it is a correctness test. For Unix I added a run-time test directory that I'm working on right now.

Copy link
Member

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

micro-benchmarks is a bad name really.. we use that directory for a whole bunch of unit tests. I'll rename it soon.

let _ = assert(fl <> "")
let fl' = string_of_list (list_of_string "Test.LexemFL.fst(11)")
let _ = assert((strlen fl') = 20) by compute()
let _ = assert(fl' = "Test.LexemFL.fst(11)") by compute()
Copy link
Member

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Using assert_norm breaks the dependency on tactics (and is maybe more idiomatic).

Copy link
Member

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Also what is this testing? Seems to be just that 1) fl is nonempty and 2) that string_of_list and list_of_string are inverses.
(Note the missing 'e' from 'Test.LexemFL.fst(11)'.)

Copy link
Contributor Author

@briangmilnes briangmilnes Oct 17, 2024

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

No what the darn string_of_list (list_of_string is doing is getting rid of oddball characters that prevent line 13 from working to show that FL is working. See the comment line above.

Copy link
Contributor Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

I have a commit pushed with the assert_norm changes. Do I need another PR?

Copy link
Contributor Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

So I have Test.FLLexeme moved to the directory to be renamed later, micro-benchmarks and validaton-time deleted and out of the makefile.

And this FStar revision will compile Luke's fl(), which is two characters shorter and more obvious than FL. I have no idea that reflection would do this for me as it is not in the book.

It would obviate this entire PR.

Guido which way do you want it? Lexeme or Tactics?

@briangmilnes
Copy link
Contributor Author

The tests/Makefile conflict is just adding validation-time as a test directory.

@mtzguido
Copy link
Member

Hi Brian, could we rename the lexeme to __FILELINE__ and rebase this PR? (rebase so the addition of validation-time is gone from the history, I guess it would be two commits: adding the lexeme, and adding the test).

It needs a fix though: we should probably take the basename of the file instead of just L.source_file. This test works for me, when running F* from the root of the repo:

module Test.LexemeFL

let fl = __FL__
let _ = assert(fl = "tests/micro-benchmarks/Test.LexemeFL.fst(3)")

(so there are no funny characters) but the behavior of the module shouldn't depend on how the file was referred to.

Lucas' tactic-based fl is really cool, and could be used for many more things, but using tactics introduces many dependencies and this new lexeme seems reasonable enough.

FWIW the syntax error you got above is probably due to an old F*. Complex meta qualifiers like the one in Lucas' code used to require enclosing parentheses.

@briangmilnes
Copy link
Contributor Author

briangmilnes commented Oct 21, 2024 via email

@briangmilnes
Copy link
Contributor Author

Too far behind, merge is not accepted on github due to some test failure, going to kill this fork and start again.

@mtzguido
Copy link
Member

For the future you can just fix up your changes locally and then force push to the branch where you opened the PR from, and it will be updated.

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Labels
None yet
Projects
None yet
Development

Successfully merging this pull request may close these issues.

3 participants