"Use of possibly unassigned field" compile error

  • Thread starter Thread starter Chris LaJoie
  • Start date Start date
C

Chris LaJoie

Our company has been developing a program in C# for some time now, and we
haven't had any problems with it, but just last night something cropped up
that has me, and everyone else, stumped.

I have a struct that contains several different types of data. This struct
is used throuout the program. Now, when I compile, I get 6 errors, all of
them "Use of possibly unassigned field 'awayTime'" or "Use of possibly
unassigned field 'intlTime'". It is correct, it is possible that these
fields aren't assigned, but I don't care. It's inside of a struct, and
everything inside of a struct gets assigned its default value anyway.

This is difficult to explain, but if anyone else has had this problem and
fixed it, i'd love to know how. Thanks.

Chris LaJoie
 
Chris LaJoie said:
Our company has been developing a program in C# for some time now, and we
haven't had any problems with it, but just last night something cropped up
that has me, and everyone else, stumped.

I have a struct that contains several different types of data. This struct
is used throuout the program. Now, when I compile, I get 6 errors, all of
them "Use of possibly unassigned field 'awayTime'" or "Use of possibly
unassigned field 'intlTime'". It is correct, it is possible that these
fields aren't assigned, but I don't care. It's inside of a struct, and
everything inside of a struct gets assigned its default value anyway.
Click to expand...

Local variables *don't* have default values, according to the C# spec.
That's what you're running into.
This is difficult to explain, but if anyone else has had this problem and
fixed it, i'd love to know how. Thanks.
Click to expand...

The easiest way to fix it is to change:

MyStruct foo;

to

MyStruct foo = new MyStruct();

in your code.
 
The struct var was being used inside of a foreach loop. Just for kicks (a
hunch) I changed the foreach to a for, looping through the array. This
solved everything.

Simply using bla = new StructName() would not have worked in this case
(because the values were being read from, not assigned to).

Chris LaJoie
 
Chris LaJoie said:
The struct var was being used inside of a foreach loop. Just for kicks (a
hunch) I changed the foreach to a for, looping through the array. This
solved everything.
Click to expand...

Interesting - could you post some sample code? (Just enough to show
what's going on.)
Simply using bla = new StructName() would not have worked in this case
(because the values were being read from, not assigned to).
Click to expand...

I'm not sure I follow you, to be honest... if you're assuming that the
declaration gave you the default values, then assigning the default
value to it at the point of declaration should give you *exactly* the
same behaviour, but without the compile error.
 
as I said earlier, my problem has been solved, but just to clear some things
up I'll post some sample code. I have my struct:

struct TestStruct {
public int a;
public int b;
}

now I have a function that only assigns the values of the struct. In my
case, this function did not get any compile errors:

public TestStruct[] Parse() {
TestStruct[] ret = new TestStruct[10];
for(int x = 0; x < ret.Length; x++) {
ret[x].a = 1;
ret[x].b = 2;
}
return ret;
}

as you can see, that function doesn't do any reading of the values, only
assigns them. I have another function, which reads the data and inserts it
in our database.

public bool InsertInDB(TestStruct[] things) {
StringBuilder sql = new StringBuilder("INSERT....");
foreach(TestStruct t in things) {
sql.Append(sqlify(t.a) + ", "); // <-- compile error here
sql.Append(sqlify(t.b) + ")"); // <-- compile error here

[...insert in db...etc...]
}
}

solving this was a matter of simply replacing the foreach with a for. I'm
not sure exactly what the difference is in the IL code generated, but
clearly there is one. here's how i fixed it:

public bool InsertInDB(TestStruct[] things) {
StringBuilder sql = new StringBuilder("INSERT....");
for(int x = 0; x < things.Length; x++) {
sql.Append(sqlify(things[x].a) + ", "); // <-- compile error here
sql.Append(sqlify(things[x].b) + ")"); // <-- compile error here

[...insert in db...etc...]
}
}

I appologize for using such silly examples, but I'm not allowed to post
company code.


Chris LaJoie
 
Chris LaJoie said:
as I said earlier, my problem has been solved, but just to clear some things
up I'll post some sample code. I have my struct:

struct TestStruct {
public int a;
public int b;
}

now I have a function that only assigns the values of the struct. In my
case, this function did not get any compile errors:

public TestStruct[] Parse() {
TestStruct[] ret = new TestStruct[10];
for(int x = 0; x < ret.Length; x++) {
ret[x].a = 1;
ret[x].b = 2;
}
return ret;
}
Click to expand...

Right - just assigning will never have any problem.
as you can see, that function doesn't do any reading of the values, only
assigns them. I have another function, which reads the data and inserts it
in our database.

public bool InsertInDB(TestStruct[] things) {
StringBuilder sql = new StringBuilder("INSERT....");
foreach(TestStruct t in things) {
sql.Append(sqlify(t.a) + ", "); // <-- compile error here
sql.Append(sqlify(t.b) + ")"); // <-- compile error here

[...insert in db...etc...]
}
}
Click to expand...

That *is* strange. I don't get a compile-time error with this, for
instance:

using System;

struct Foo
{
public int x;
}

class Test
{
static void Main(string[] args)
{
Foo[] f = new Foo[5];
f[0].x=5;
Bar (f);
}

static void Bar (Foo[] x)
{
foreach (Foo y in x)
{
Console.WriteLine (y.x);
}
}
}

I appologize for using such silly examples, but I'm not allowed to post
company code.
Click to expand...

That's not a problem at all - but I'm still finding it hard to see why
you're getting an error in the first place. Could you try compiling my
code above? It should compile without warning. If it does, could you
try constructing a similar complete test case which *doesn't* compile?

Which version of Visual Studio (or csc) are you using?
 
I'm using VS.NET 2003. If I tried to create an example program that didn't
compile, I wouldn't be able to. I honestly believe that this is a problem
with the C# compiler. These problems crop up in code that HAS worked for
months. Usually silly things like changing a foreach to a for will fix it.
For example, I fixed one a couple weeks ago. Only one member of the struct
was causing compile errors, so I changed the code around inside the struct.
Here's how I did it.

struct TestStruct {
public int a; // <-- still works fine
//public int b; // <-- broken
private int m_b;
public int b {
get {return m_b;}
set {m_b = value;}
}
}

that's right, simply changing a public variable to a property fixed the
problem.

I've researched the error (Use of possibly unassigned field) thoroughly.
Google only came up with documentation, not anyone else with problems.
Something I haven't tried is to compile it on another computer when it does
not compile on mine.

I do appreciate your interest in this, but I'm afraid that in the end we
won't be any closer to figuring it out.

Chris LaJoie

Jon Skeet said:
Chris LaJoie said:
as I said earlier, my problem has been solved, but just to clear some things
up I'll post some sample code. I have my struct:

struct TestStruct {
public int a;
public int b;
}

now I have a function that only assigns the values of the struct. In my
case, this function did not get any compile errors:

public TestStruct[] Parse() {
TestStruct[] ret = new TestStruct[10];
for(int x = 0; x < ret.Length; x++) {
ret[x].a = 1;
ret[x].b = 2;
}
return ret;
}
Click to expand...

Right - just assigning will never have any problem.
as you can see, that function doesn't do any reading of the values, only
assigns them. I have another function, which reads the data and inserts it
in our database.

public bool InsertInDB(TestStruct[] things) {
StringBuilder sql = new StringBuilder("INSERT....");
foreach(TestStruct t in things) {
sql.Append(sqlify(t.a) + ", "); // <-- compile error here
sql.Append(sqlify(t.b) + ")"); // <-- compile error here

[...insert in db...etc...]
}
}
Click to expand...

That *is* strange. I don't get a compile-time error with this, for
instance:

using System;

struct Foo
{
public int x;
}

class Test
{
static void Main(string[] args)
{
Foo[] f = new Foo[5];
f[0].x=5;
Bar (f);
}

static void Bar (Foo[] x)
{
foreach (Foo y in x)
{
Console.WriteLine (y.x);
}
}
}

I appologize for using such silly examples, but I'm not allowed to post
company code.
Click to expand...

That's not a problem at all - but I'm still finding it hard to see why
you're getting an error in the first place. Could you try compiling my
code above? It should compile without warning. If it does, could you
try constructing a similar complete test case which *doesn't* compile?

Which version of Visual Studio (or csc) are you using?
Click to expand...
 
Chris LaJoie said:
I'm using VS.NET 2003. If I tried to create an example program that didn't
compile, I wouldn't be able to. I honestly believe that this is a problem
with the C# compiler.
Click to expand...

Yes, it sounds like it. Certainly the code you posted looked like it
was fine.
These problems crop up in code that HAS worked for
months. Usually silly things like changing a foreach to a for will fix it.
For example, I fixed one a couple weeks ago. Only one member of the struct
was causing compile errors, so I changed the code around inside the struct.
Click to expand...

<snip>

That sounds very strange.
Something I haven't tried is to compile it on another computer when it does
not compile on mine.
Click to expand...

That's definitely worth doing. I can't imagine how your VS.NET
installation could be damaged in such a subtle way, but it's definitely
worth a try.
I do appreciate your interest in this, but I'm afraid that in the end we
won't be any closer to figuring it out.
Click to expand...

You could well be right :(
 
Ah Hah!

As it turns out this is a known bug in the 7.0 and 7.1 C# compiler. It is
caused by having semi-large structs or just a large number of structs as
locals. There really is no good work-around other than to try and change
the method to decrease the number of structs, or other locals. Sometimes
even increasing them will also work. This has been fixed in current builds.
Those of you lucky enough to get bits from PDC can verify the fix if you
feel like it. I am working with the 'powers that be' to see if this fix can
also get pushed out as part of a future service pack or something, but no
promises!
 
Back
Top