[webkit-efl] Symbols visibility

[Lucas De Marchi]

Hi Tomasz

On Tue, Feb 28, 2012 at 10:18 AM, Tomasz Morawski
<t.morawski at samsung.com> wrote:
> Hi Lucas De Marchi,
>
>> We started to use? It's there since almost 2 years ago.
>
> OK, my wrong I have used wrong words. I was a little in shock when I
> found how it works. I asked few people which work with this port about "how
> to export function from from efl port" and everybody said that I should use
> EAPI. Anyway, it is not "almost 2" years also. ;)

Humn... my bad, I thought I had implemented it earlier.

>
>
>> We need to export this "c++ symbol" because of DRT; nothing normal
>> users should care about.
>
> Yes, I know that you have need it for DRT tool. So, we have public API
> layer and "hidden" API layer for internal WebKit tools. Does it true?

No, it gets exported as well.

>
>
>>> how we will control symbols visiblity. I think linker
>>
>> I'm not saying it cannot be revisited, but this was discussed 2 years
>> ago. See the corresponding bug in bugzilla, I remember I put some
>> numbers there. The thing is that we try to hide all symbols that we
>> can.
>
> How it was before linker script? Did visibility ("hidden") or
> -fvisibility=hidden was used before?

We can't use -fvisibility=hidden because it breaks build in case we
use SHARED_CORE. A bunch of unneeded symbols were being exported:
https://bugs.webkit.org/show_bug.cgi?id=44609

>
>
>> Have you seen  Source/autotools/symbols.filter ?
>
> Yes, I have. Why are you asking?

Because you said no other port uses linker script. GTK clearly does.

> I would like to discuses this topic only how we should clean up EAPI mess
> now or left it as it it. I am asking about it due to current situation is
> not clear:
>
> I would like to ask:
> 1) what EAPI macro means in our port now?

It remembers you that function is exported and allows you to easily grep code.

> 2) how to find out base on code which function/c++ object is exported?

No C++ object should be exported. The only exception is the one needed for DRT.


> 3) Should we not use EAPI any longer? Should we delete EAPI from our
> headers?

IMO, just let it as is.


Lucas De Marchi