Note that usages of NonEmptyValidation inherited classes should be checked as well:
BaseI18NValidation
ByteSizeValidation
IpAddressValidation
MacAddressValidation
NoSpacesValidation

Note that generally, doesn't matter if empty string is either legal or illegal -  it should be encapsulated within the regular expression itself.
So in case there has been NonEmptyValidation in use in addition to the RegexValidation, need to make sure that the regular-expression by itself blocks non-empty strings. If not - it should be changed accordingly (assuming it is easy enough to do).

in gerrit: http://gerrit.ovirt.org/#change,4690

correction:

(In reply to comment #0)
> - Need to remove the unnecessary NotEmptyValidation usages (practically in
> every place in the code in which RegexValidation is being used)

no need to do that; using NotEmptyValidation results in a dedicated validation error message in case of an empty string ("This field can't be empty"), which is nicer.
However, In order to make sure that this message will appear for empty values, it should be passed first in the IValidation array passed to the validateEntity() method. That way it will be processed first.
One way to avoid making sure that the NotEmptyValidation is processed first is to pass a boolean parameter to "validateEntity()" - either "emptyValueLegal" or "processEmptyValue"
- when passing to validateEntity() an empty value and "emptyValueLegal == false" (or "processEmptyValue == false"), validateEntity() should return "false" immediately and return the "this field can't be empty" message.
- when passing to validateEntity() an empty string and "emptyValueLegal == true" (or "processEmptyValue == true"), validateEntity() should return "true" immediately (or simply process the empty value regularly (according to the IValidation array)).

> - Error messages for RegexValidation should be rephrased accordingly.
> I.e., a message like:
>  "Name must contain alphanumeric characters only." 
> should become:
>  "Name must be non-empty and contain alphanumeric characters only."

Should still be done, in order to keep the integrity / full match of regular expression to validation error message.

> One way to avoid making sure that the NotEmptyValidation is processed first is > to pass a boolean parameter to "validateEntity()" - either "emptyValueLegal" or > "processEmptyValue"
That would mean, that we will change from declarative style like this:
someField.ValidateEntity(
   new IValidation[] {
   new NotEmptyValidation(),
   new RegexValidation(nameExpr, nameMsg)
});
... 

to procedural like this:
...
someField.ValidateEntity(false,
  new IValidation[] {
  new RegexValidation(nameExpr, nameMsg)
});
....

I would say, that the first possibility is more readable than the second (e.g. do you know what does the "false" means without opening the ValidateEntity method?) 

Also the second possibility would mix up two things:
- EntityModel.ValidateEntity would still be an "engine" which orchestrates the validations
- EntityModel.ValidateEntity would become sometimes also a validation (depending on a boolean switch).

I would say, that the instances of IValidation should do one-purpose validations (like LengthValidation, IntegerValidation, NotEmptyValidation, LinuxMountPointValidation...)
and should be passed in the correct order to the EntityModel.ValidateEntity which will orchestrate them. It separates the concerns clearly and is (considering the validations are well named) readable.

Closing old bugs. If this issue is still relevant/important in current version, please re-open the bug.