Smalltalk › Squeak › Squeak VM

longRunningPrimitiveCheckSequenceNumber compiler warning

_‹ Previous Topic Next Topic _›

Classic

List

Threaded

2 messages Options

Nicolas Cellier

longRunningPrimitiveCheckSequenceNumber compiler warning

We have this code

voidLongRunningPrimitive: reason

...

    <inline: false>
    longRunningPrimitiveCheckMethod := nil.
    longRunningPrimitiveCheckSequenceNumber = 0. "not strictly necessary but prevents this being inlined into checkForLongRunningPrimitive"
   ...

which generates

static void NoDbgRegParms
voidLongRunningPrimitive(char *reason)
{   DECL_MAYBE_SQ_GLOBAL_STRUCT
    GIV(longRunningPrimitiveCheckMethod) = null;
    GIV(longRunningPrimitiveCheckSequenceNumber) == 0;

...

and a compiler warning suggesting to replace == with =

So it seems to effectively be a hack, but it seems fragile.

Couldn't we remove this line (don't we have inline: #never for that purpose?).

Eliot Miranda-2

Re: longRunningPrimitiveCheckSequenceNumber compiler warning

Hi Nicolas,

On Fri, Jan 4, 2019 at 3:39 AM Nicolas Cellier <[hidden email]> wrote:

We have this code

voidLongRunningPrimitive: reason
    ...
    <inline: false>
    longRunningPrimitiveCheckMethod := nil.
    longRunningPrimitiveCheckSequenceNumber = 0. "not strictly necessary but prevents this being inlined into checkForLongRunningPrimitive"
   ...

which generates

static void NoDbgRegParms
voidLongRunningPrimitive(char *reason)
{   DECL_MAYBE_SQ_GLOBAL_STRUCT
    GIV(longRunningPrimitiveCheckMethod) = null;
    GIV(longRunningPrimitiveCheckSequenceNumber) == 0;
   ...

and a compiler warning suggesting to replace == with =
So it seems to effectively be a hack, but it seems fragile.
Couldn't we remove this line (don't we have inline: #never for that purpose?).

Yes. When I wrote the code we didn't have inline: #never. inline: #never is a much better solution.

_,,,^..^,,,_

best, Eliot