%subclass should provide an @interface

[DHowett]

Right now, %subclass treats self as an id. Logos could create a secret @interface (_logos_interface$_ungrouped$Blah, maybe?) and use that for self.

[jithatsonei]

Is this already a thing @kirb?

[kirb]

No, and I'm not entirely sure how to go about this because it would break code that already has its own definition of the @interface. Maybe one day we'll get into semantic versioning so we can make breaking changes like this one (#111).

[DHowett]

If Logos could detect that (to wit: become an actual objc parser), we could
use the user-provided class type or generate one as necessary.
On Feb 14, 2016 12:46 AM, "Adam Demasi" [email protected] wrote:

No, and I'm not entirely sure how to go about this because it would break
code that already has its own definition of the @interface. Maybe one day
we'll get into semantic versioning so we can make breaking changes like
this one (#111).

On Sun, 14 Feb 2016 at 14:07, alucard0134 [email protected]
wrote:

Is this already a thing @kirb https://github.com/kirb?

—
Reply to this email directly or view it on GitHub
https://github.com/theos/theos/issues/100#issuecomment-183815426.

Adam Demasi
HASHBANG Productions
hbang.com.au

—
Reply to this email directly or view it on GitHub
https://github.com/theos/theos/issues/100#issuecomment-183853098.

[kirb]

I'm a bit reluctant on really changing Logos much because I'm considering whether it could be replaced by @eswick's Logos-ish clang preprocessor (also, I'm terrible at Perl). Of course that introduces the need to compile code, and quite a lot of code at that. (Will make another issue for this)

(Edit: Lesson learned: email response does not parse Markdown 👎)

[DHowett]

Problem there is, it hasn't kept up with Clang as it's changed. I'm worried
that it ties people to whatever arbitrary version of whatever random
compiler they built when they got Theos.
On Feb 14, 2016 12:55 AM, "Adam Demasi" [email protected] wrote:

I'm a bit reluctant on really changing Logos much because I'm considering
whether it could be replaced by @eswick's Logos-ish clang preprocessor (also, I'm terrible at Perl). Of
course that introduces the need to compile code, and quite a lot of code at
that. (Will make another issue for this)

On Sun, 14 Feb 2016 at 19:18, Dustin L. Howett [email protected]
wrote:

If Logos could detect that (to wit: become an actual objc parser), we
could
use the user-provided class type or generate one as necessary.
On Feb 14, 2016 12:46 AM, "Adam Demasi" [email protected]
wrote:

No, and I'm not entirely sure how to go about this because it would
break
code that already has its own definition of the @interface. Maybe one
day
we'll get into semantic versioning so we can make breaking changes like
this one (#111).

On Sun, 14 Feb 2016 at 14:07, alucard0134 [email protected]
wrote:

Is this already a thing @kirb https://github.com/kirb?

—
Reply to this email directly or view it on GitHub
https://github.com/theos/theos/issues/100#issuecomment-183815426.

Adam Demasi
HASHBANG Productions
hbang.com.au

—
Reply to this email directly or view it on GitHub
https://github.com/theos/theos/issues/100#issuecomment-183853098.

—
Reply to this email directly or view it on GitHub
https://github.com/theos/theos/issues/100#issuecomment-183853280.

Adam Demasi
HASHBANG Productions
hbang.com.au

—
Reply to this email directly or view it on GitHub
https://github.com/theos/theos/issues/100#issuecomment-183855051.

[kirb]

He also has a conflict of interest (recent Apple internship) so it may be infeasible anyway from a maintenance standpoint unless someone else can take over.

[eswick]

I recconend sticking with Perl. With Clang, we'd have to update code for every new version. Like you said, I have an internship at Apple and don't have the time nor the will to keep it updated.

[andrewwiik]

Um I know this isn't the place but have you guys ever though of making a way to de-initlize groups?

[rpetrich]

@andrewwiik: It isn't possible to implement that in a safe way with all of the behaviours that developers expect. Instead, it can be done in the tweak itself: always call %init() at startup and check a shouldPlayDead variable in the method hooks.

[kirb]

Please create a new issue in future or get in touch on IRC or Twitter rather than hijacking an existing thread, @andrewwiik.