Caution when using Next Repeat

[Simon Knight]

I have just wasted a morning trying to discover why the following code failed:

Code: Select all

if (there is file tNewFilePath) then
      -- skip this file
      put "New name already exists, so skipping new name  : " & tNewFileName & cr after tErrorsLog
      add one to tErrorCount
      next repeat
   end if

The snip was part of a file renaming handler that would run for thousands of files and then stop with an error in the IDE. Several things were odd: the error message in the IDE claimed there was a problem with Character 1 of the first line (if there is a file tNewFilePath then), when wrapped in a Try Catch End Try the IDE reported the error on wherever the try statement was placed above the code snip (note there were quite a few lines before the code snip) and the catch statement did not catch any error.

After much head scratching I realised the issue was with the line "next repeat". This line is an orphan as I removed the original repeat loop and failed to notice the line. The error only occurred when the if statement was true which was a rare event and the reason why the rest of the code processed thousands of files before failing.

I think the IDE should post a syntax error and not allow the code to be applied much as it does with if statements with no end ifs - what do you think?

S

[stam]

I agree the IDE should post a warning if a next repeat statement is used outside a of a repeat loop... may be worth posting a bug report/quality improvement request.

[mwieder]

Definitely bug-worthy behavior.
I'm not even sure what to expect when an orphan next repeat command would get executed.

[Simon Knight]

"I'm not even sure what to expect when an orphan next repeat command would get executed."

A slightly unhelpful error occurs in the IDE at the start of the block that contains the line. I have not tried it in a built application but I suspect that it will be a silent fail as no error is trapped within a try, catch,end try code block.

S

[FourthWorld]

A slightly unhelpful error occurs in the IDE at the start of the block that contains the line. I have not tried it in a built application but I suspect that it will be a silent fail as no error is trapped within a try, catch,end try code block.

Errors are reported by the engine in at least three different ways, depending on the nature of the error.

What is the message you see in the IDE, and where does the IDE present this message?

[mwieder]

So... interestingly, the code compiles with no problems (no context checking).
Executing this code does throw an error at the 'next repeat' line.
Placing the 'next repeat' in a try/catch construct throws the same error but earmarks it at the enclosing 'try' line.
And placing that in an if construct moves the error report to the outermost conditional.

Code: Select all

local tVariable
   if tVariable is empty then
      try
         next repeat
      catch e
         answer e
      end try
   end if

In an ideal world (looking at you, 10.0) the editor would catch this in the edit phase, the way it does if I enter "end repeat" instead of "next repeat".

[stam]

It seems like something that should be reported by the IDE before run time.
All good and well hoping that v10.0 will fix this, but if a bug report doesn't exist for this already maybe you could submit a bug/feature request?

[Simon Knight]

Having prepared a demo stack for my bug report I did the search of the quality control site. The issue was first reported in 2015 and has been re-reported several times since. Seems that there are also issues with exit as well.

Anyhow I attach my demo stack so that others may enjoy the problem

S

NextRepeatError.livecode.zip

Demo of orphan Next Repeat failure

(1.59 KiB) Downloaded 65 times

[FourthWorld]

Executing this code does throw an error at the 'next repeat' line.

What is the error message?

[Simon Knight]

The only message I saw was error at line n character 1.

S

[FourthWorld]

The only message I saw was error at line n character 1.

Was n the correct line?

Vague, yes, and I'd love to see more Clang-like error assistance. But it's useful as a starting point to assess if the engine is at least covering the basics.

[Simon Knight]

Hi Richard,

Was line n correct ? No. The code stops in the IDE on the top level structure, meaning that if there is a TRY statement it stops and reports the line the Try is on. Try my demo stack above to see. In effect the code executes down to the line "next repeat" then jumps up to the Try or the If statement and reports the error.

S

[FourthWorld]

Hi Richard,

Was line n correct ? No. The code stops in the IDE on the top level structure, meaning that if there is a TRY statement it stops and reports the line the Try is on.

The try control structure alters much with exception handing, by its nature. I suppose it could be further refined to provide more context sensitivity, but the inner workings of try seem tangential to the core concern here about orphaned loop closures.

In the original code, before try was attempted as a remedy, was n the appropriate line number?

[Simon Knight]

Hi Richard,

Short answer : No.

Long answer, please down load my demo stack which is posted above which will answer many of your questions. Suffice to say the error line reported was never the problem line. When the line "Next return" was within an IF structure the overall app would run until the line was actioned based on the IF structure. When hit the error was incorrectly reported it as occurring on the line of the IF statement that contained the line "Next Return". When wrapped in a TRY CATCH the error is reported on the line of the TRY statement.

S

[mwieder]

...and the reported error is 241: "error in statement". Can't get much more generic than that.
The more glaring problem, I think, is reporting the error on the outermost conditional.