Bug 76702 – [GTK] Detect cross-compilation in webkitdirs.pm to not generate gtkdoc

• Status: RESOLVED
• Resolution: FIXED
• Component: Tools / Tests

[Gabor Rapcsanyi]

It would be good if we can avoid the gtkdoc generating if it doesn't need for us.
For example it would make easier the cross-compilation.

[Gabor Rapcsanyi]

Created attachment 123302 [details]
proposed patch

[Martin Robinson]

Comment on attachment 123302 [details]
proposed patch

Perhaps it should be a command-line option instead? Out of curiosity, why do you want to avoid it?

[Gabor Rapcsanyi]

(In reply to comment #2)
> (From update of attachment 123302 [details])
> Perhaps it should be a command-line option instead? Out of curiosity, why do you want to avoid it?

I've just checked webkitdirs.pm and saw a lot environment variables thats why I chose this way, but if you prefer a command-line option for build-webkit then I can make it as well.
Anyway its need for GtkWebKit cross-compilation for ARM, and if you want to compile on x86 you can't run the ARM binaries for gtkdoc. Thats why we want to avoid this.

[Martin Robinson]

(In reply to comment #3)

> I've just checked webkitdirs.pm and saw a lot environment variables thats why I chose this way, but if you prefer a command-line option for build-webkit then I can make it as well.
> Anyway its need for GtkWebKit cross-compilation for ARM, and if you want to compile on x86 you can't run the ARM binaries for gtkdoc. Thats why we want to avoid this.

In that case it might be better just to detect situations in which we are cross-compiling and avoid gtkdoc there.

[Gabor Rapcsanyi]

(In reply to comment #4)
> (In reply to comment #3)
> 
> > I've just checked webkitdirs.pm and saw a lot environment variables thats why I chose this way, but if you prefer a command-line option for build-webkit then I can make it as well.
> > Anyway its need for GtkWebKit cross-compilation for ARM, and if you want to compile on x86 you can't run the ARM binaries for gtkdoc. Thats why we want to avoid this.
> 
> In that case it might be better just to detect situations in which we are cross-compiling and avoid gtkdoc there.

But what if somebody don't want gtkdoc, why don't we give an option or something to avoid this? For example we don't need it on our bot we just need the binaries. In other hand I think its simpler than detect if somebody trying to cross-compile.

[Martin Robinson]

(In reply to comment #5)

> But what if somebody don't want gtkdoc, why don't we give an option or something to avoid this? For example we don't need it on our bot we just need the binaries. In other hand I think its simpler than detect if somebody trying to cross-compile.

I think it's good that build-webkit just does the right thing with minimal intervention.

[Martin Robinson]

(In reply to comment #6)
> (In reply to comment #5)
> 
> > But what if somebody don't want gtkdoc, why don't we give an option or something to avoid this? For example we don't need it on our bot we just need the binaries. In other hand I think its simpler than detect if somebody trying to cross-compile.
> 
> I think it's good that build-webkit just does the right thing with minimal intervention.

I should also mention that documentation build failures are now considered build failures. That's why we are building documentation while building the library.

[Gabor Rapcsanyi]

Created attachment 123932 [details]
proposed patch v2

I have made an isCrossCompilation() method to webkitdirs.pm. It detects whether the compiler is configured with different host and target options. So now the script can determine to generate gtkdoc or not.

[Gyuyoung Kim]

Comment on attachment 123932 [details]
proposed patch v2

Attachment 123932 [details] did not pass efl-ews (efl):
Output: http://queues.webkit.org/results/11159528

[Early Warning System Bot]

Comment on attachment 123932 [details]
proposed patch v2

Attachment 123932 [details] did not pass qt-ews (qt):
Output: http://queues.webkit.org/results/11342523

[Gabor Rapcsanyi]

Created attachment 123934 [details]
proposed patch v3

[Martin Robinson]

Comment on attachment 123934 [details]
proposed patch v3

Thanks!

[WebKit Review Bot]

Comment on attachment 123934 [details]
proposed patch v3

Rejecting attachment 123934 [details] from commit-queue.

New failing tests:
media/audio-garbage-collect.html
Full output: http://queues.webkit.org/results/11294581

[Gabor Rapcsanyi]

Landed by: http://trac.webkit.org/changeset/105974

[thouraya]

Hi,

It is causing  build issue for sh4 platforms due to the check for "--host" in compiler_options.

the "--host" flags is not needed for a cross-compiler?


Regards,
Thouraya.



(In reply to comment #14)
> Landed by: http://trac.webkit.org/changeset/105974

[Gabor Rapcsanyi]

(In reply to comment #15)
> 
> Hi,
> 
> It is causing  build issue for sh4 platforms due to the check for "--host" in compiler_options.
> 
> the "--host" flags is not needed for a cross-compiler?
> 
> 
> Regards,
> Thouraya.
> 
> 
> 
> (In reply to comment #14)
> > Landed by: http://trac.webkit.org/changeset/105974

This isCrossCompilation() function is just checking the --target and --host flags of the compiler. If they are set and differ then we return true. Any other case we return with false. So they need to be set.

[thouraya]

> This isCrossCompilation() function is just checking the --target and --host >flags of the compiler. If they are set and differ then we return true. Any other >case we return with false. So they need to be set.


Is it possible to check only the --target flag?

[Gabor Rapcsanyi]

(In reply to comment #17)
> > This isCrossCompilation() function is just checking the --target and --host >flags of the compiler. If they are set and differ then we return true. Any other >case we return with false. So they need to be set.
> 
> 
> Is it possible to check only the --target flag?

Unfortunately it's not enough because if the target is ARM you can't just say that now we detected a cross compiler because if the host is ARM as well but there is no --host flag then it is native building not cross compilation.
What is the exact problem on your side?

[thouraya]

Hi,

> What is the exact problem on your side?

The compiler I'm using is not configured with the option (--host)

and recompiling it with the option --host may impact other tools.

Regards,
Thouraya.

[Gabor Rapcsanyi]

(In reply to comment #19)
> Hi,
> 
> > What is the exact problem on your side?
> 
> The compiler I'm using is not configured with the option (--host)
> 
> and recompiling it with the option --host may impact other tools.
> 
> Regards,
> Thouraya.

Hmm, then maybe you can make a bug for this problem and add some extra check to this function. For example the 'arch' command could helps you out or just introduce an environment variable for that.