Currently, std.string.assumeUTF is inferred @nogc in release mode, while in debug mode it isn't. Should this be allowed?
It can cause some problems, e.g. I have some @nogc unittests that I use to guarantee that my code is in fact @nogc, and these fail because in debug mode assumeUTF is not.
In general (IMHO) this kind of difference should be avoided, because code should not change meaning between debug and release modes.

What do you think?

On Thursday, 7 July 2016 at 19:39:12 UTC, Lodovico Giaretta wrote:
> Currently, std.string.assumeUTF is inferred @nogc in release mode, while in debug mode it isn't. Should this be allowed?
> It can cause some problems, e.g. I have some @nogc unittests that I use to guarantee that my code is in fact @nogc, and these fail because in debug mode assumeUTF is not.
> In general (IMHO) this kind of difference should be avoided, because code should not change meaning between debug and release modes.
>
> What do you think?

That is definetly a bug.

On 7/8/2016 1:04 PM, Jerry wrote:
> On Thursday, 7 July 2016 at 19:39:12 UTC, Lodovico Giaretta wrote:
>> Currently, std.string.assumeUTF is inferred @nogc in release mode, while in
>> debug mode it isn't. Should this be allowed?
>> It can cause some problems, e.g. I have some @nogc unittests that I use to
>> guarantee that my code is in fact @nogc, and these fail because in debug mode
>> assumeUTF is not.
>> In general (IMHO) this kind of difference should be avoided, because code
>> should not change meaning between debug and release modes.
>>
>> What do you think?
>
> That is definetly a bug.

Not a bug until it is posted to Bugzilla!

On Saturday, 9 July 2016 at 01:05:54 UTC, Walter Bright wrote:
> Not a bug until it is posted to Bugzilla!

so if i won't post a dmd segfault there, i may assume that dmd segfault is intended behavior? then let's document it! ;-)

On Saturday, 9 July 2016 at 01:05:54 UTC, Walter Bright wrote:
> On 7/8/2016 1:04 PM, Jerry wrote:
>> That is definetly a bug.
>
> Not a bug until it is posted to Bugzilla!

Well, I didn't post it there directly because I wasn't sure it could be considered a bug. The reason it is not @nogc in debug mode is that it performs extra checks that may throw exceptions (so I guess it isn't nothrow either), which isn't bad per se, but makes it difficult to verify (using unittest) that code using it is @nogc.