www.digitalmars.com         C & C++   DMDScript  

digitalmars.D.bugs - Dependent version assignments only conditionally work (DMD 0.121)

reply Burton Radons <burton-radons smocky.com> writes:
This code succeeds compilation:

    version (b) illegal declaration;
    version (a) version = b;
    version = a;

However, it should fail because version "a" should be defined, then 
version "b", then expanding the illegal declaration.  This correctly fails:

    version (b) illegal declaration;
    version = a;
    version (a) version = b;

The compiler is only doing one pass through the declarations to find 
versions to assign; instead it should restart the loop anytime it adds a 
new version or debug.
Apr 17 2005
next sibling parent Thomas Kuehne <thomas-dloop kuehne.thisisspam.cn> writes:
-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA1

Burton Radons schrieb am Sun, 17 Apr 2005 17:47:45 -0700:
 This code succeeds compilation:

     version (b) illegal declaration;
     version (a) version = b;
     version = a;

 However, it should fail because version "a" should be defined, then 
 version "b", then expanding the illegal declaration.  This correctly fails:

     version (b) illegal declaration;
     version = a;
     version (a) version = b;

 The compiler is only doing one pass through the declarations to find 
 versions to assign; instead it should restart the loop anytime it adds a 
 new version or debug.

Added to DStress as http://dstress.kuehne.cn/run/debug_08.d http://dstress.kuehne.cn/run/debug_09.d http://dstress.kuehne.cn/run/debug_10.d http://dstress.kuehne.cn/run/version_33.d http://dstress.kuehne.cn/run/version_34.d http://dstress.kuehne.cn/run/version_35.d Thomas -----BEGIN PGP SIGNATURE----- iD8DBQFCY1e13w+/yD4P9tIRAjOpAJ90EP64D2Z002Ny6mrUOaEx+JPAzwCdFOMN gkvkE4pXDDGhV7xvhhAegWI= =ppeG -----END PGP SIGNATURE-----
Apr 17 2005
prev sibling parent reply "Walter" <newshound digitalmars.com> writes:
"Burton Radons" <burton-radons smocky.com> wrote in message
news:d3v03l$2oag$1 digitaldaemon.com...
 This code succeeds compilation:

     version (b) illegal declaration;
     version (a) version = b;
     version = a;

 However, it should fail because version "a" should be defined, then
 version "b", then expanding the illegal declaration.  This correctly

     version (b) illegal declaration;
     version = a;
     version (a) version = b;

 The compiler is only doing one pass through the declarations to find
 versions to assign; instead it should restart the loop anytime it adds a
 new version or debug.

I don't agree. Since version assignments are only local to a module, there's no need to support forward references of versioning, and in fact, one could get into trouble with things like: version (a) version = b; version (b) version = a; There just seems no point to supporting forward referencing of versions.
Apr 18 2005
next sibling parent Thomas Kuehne <thomas-dloop kuehne.thisisspam.cn> writes:
-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA1

Walter schrieb am Mon, 18 Apr 2005 22:46:26 -0700:
 "Burton Radons" <burton-radons smocky.com> wrote in message
 news:d3v03l$2oag$1 digitaldaemon.com...
 This code succeeds compilation:

     version (b) illegal declaration;
     version (a) version = b;
     version = a;

 However, it should fail because version "a" should be defined, then
 version "b", then expanding the illegal declaration.  This correctly

     version (b) illegal declaration;
     version = a;
     version (a) version = b;

 The compiler is only doing one pass through the declarations to find
 versions to assign; instead it should restart the loop anytime it adds a
 new version or debug.

I don't agree. Since version assignments are only local to a module, there's no need to support forward references of versioning, and in fact, one could get into trouble with things like: version (a) version = b; version (b) version = a; There just seems no point to supporting forward referencing of versions.

Having complex feature<->debug dependencies within a module can require forward referencing. There is no risk of infinit loops since version/debug can only be set. Setting a version/debug if it is allready set shouldn't trigger a restart of the loop. Thomas -----BEGIN PGP SIGNATURE----- iD4DBQFCZPbL3w+/yD4P9tIRAkscAJiPb+rX+/DG8hiFVq9w/oRqYt4HAJ4mN2mW ZXtw0yKAcvHsjcaVnFfOrw== =FCzQ -----END PGP SIGNATURE-----
Apr 19 2005
prev sibling parent Burton Radons <burton-radons smocky.com> writes:
Walter wrote:

 "Burton Radons" <burton-radons smocky.com> wrote in message
 news:d3v03l$2oag$1 digitaldaemon.com...
 
This code succeeds compilation:

    version (b) illegal declaration;
    version (a) version = b;
    version = a;

However, it should fail because version "a" should be defined, then
version "b", then expanding the illegal declaration.  This correctly

fails:
    version (b) illegal declaration;
    version = a;
    version (a) version = b;

The compiler is only doing one pass through the declarations to find
versions to assign; instead it should restart the loop anytime it adds a
new version or debug.

I don't agree. Since version assignments are only local to a module, there's no need to support forward references of versioning, and in fact, one could get into trouble with things like: version (a) version = b; version (b) version = a; There just seems no point to supporting forward referencing of versions.

Err, so why did you support them then? :) version (a) illegal declaration; version = a; This works, as in not works. I'm just asking that the compiler behave consistently. If forward version referencing weren't allowed altogether, I'd be fine with that. Your example wouldn't do anything. If either version a or b were enabled, then it would enable the other, opening the other statement, then do nothing because the original is already enabled. We can't get into loops with version identifiers because we can't disable them, but with version numbers there's a potential (start with version = 1): version (1) version = 0; else version = 1; More than one version number in a module is impossible to evaluate without taking some order dependency so could safely be considered illegal. In fact, if forward version referencing is allowed it might be best to make them allowed but they must always increase the version.
Apr 19 2005