Use of unassigned vars

[FFF]

Folks,

Code: Select all

TRY
Response := (HttpWebResponse)Request:GetResponse()
 SR := StreamReader{Response:GetResponseStream()}
	ResponseData := SR:ReadToEnd()
CATCH Wex AS System.Net.WebException
 SR := StreamReader{Wex.Response:GetResponseStream()}
	ResponseData := SR:ReadToEnd()
THROW Exception{ResponseData}
FINALLY
 SR:Close() // line 61
END TRY

gives me a
warning XS0165: Use of unassigned local variable 'SR' 61,3 EasyHttp.prg
Why?


Karl

[wriedmann]

Hi Karl,

you are in a Try/Catch block, and if your code fails with an exception other than the one you are catching, before SR is initialized, you call the :Close() method on a not initialized object.
This sort of compiler warning helps me a lot!

Wolfgang

[FFF]

Wolfgang,
yep, i hear you.
But what to do? As far as i saw, i have to provide the input stream at instantiation of the object, so when moving instantiation out of the block i'm unsafe, in case the stream call fails. OTOH, reminds me why i don't like the concept of writing constructors which do more than producing an object..
Somewhat contradictory to the "assign vals to property in the constructor call in the other thread.
BTW, i tried the new conditional syntax oSR?:Close() to no avail, not sure what to make of that..

Finally, in my defense: this is not my code, but the transposition of VB source

Thx
Karl

[wriedmann]

Hi Karl,

I would set the SR to null before the Try statement, so the error goes away. And in the Finally block I would add a "if SR != null" statement.

It seems X# makes better warnings that VB.NET - IMHO the warning is 100% correct.

Wolfgang

[FFF]

Hi Wolfgang,
not sure what use a := Null has? If both instantiation paths in try and catch fail, the final :close would crash, nevertheless.
I think the ?: syntax prevents a crash...

But i wonder: i get no warning in the ResponseData := SR:ReadToEnd() lines, probably because the instantiation is "inside" the same block. What means, the real warning occurs because the compiler can't feather that SR exists as either the try or the catch path WILL be executed.

In a certain way the inverse situation of "unreachable code"

Not sure if this qualifies as a bug or not...
Anyway, interesting discussion,
thx for you time!
Karl

[wriedmann]

Hi Karl,

personally I use statements like

Code: Select all

SR := null

often also in my VO applications to keep the compiler calm, and to say explicetely that I know that this variable can be null.

Wolfgang

[Frank Maraite]

Hi Karl,
in your first line after try you have

Code: Select all

Response := (HttpWebResponse)Request:GetResponse()

This may fail to another then a System.Net.WebException. This (other) exception is not catched. So the second line

Code: Select all

SR := StreamReader{Response:GetResponseStream()}

is not executed and not in any catch block.

Solution:

Code: Select all

LOCAL SR := NULL AS StreamReader
..
FINALLY
IF SR != NULL
   SR:Close()
ENDIF

I experience that the compiler is always right with that warning. It reminds you that SR may be uninitialized without intend and you have to check out this case.

Frank