Bug 30342 – Eliminate the use of String::format

• Status: NEW
• Component: Text

[Evan Martin]

String::format relies internally on printf(), which is in turn affected by the system locale in how it handles decimal separators, etc.

[Evan Martin]

Created attachment 41129 [details]
patch written by Alp Toker

In https://bugs.webkit.org/show_bug.cgi?id=18994 , Darin Adler r-'d this patch, but it is a step in the right direction.

[Dirk Schulze]

*** Bug 31514 has been marked as a duplicate of this bug. ***

[Dirk Schulze]

*** Bug 18985 has been marked as a duplicate of this bug. ***

[Ademar Reis]

Please don't forget of WebKit2, String::format() is used, for example, in the LocalizationStrategy:

WebKit2/WebProcess/WebCoreSupport/WebPlatformStrategies.cpp

[Darin Adler]

We’ve reduced the number of call sites using String::format, but there are still quite a few left.

[Michael Catanzaro]

Darin, was the dependency on bug #176305 a mistake? It looks unrelated?

[Darin Adler]

(In reply to Michael Catanzaro from comment #6)
> Darin, was the dependency on bug #176305 a mistake? It looks unrelated?

Yes, typo in bug number.

[Darin Adler]

(In reply to Darin Adler from comment #7)
> (In reply to Michael Catanzaro from comment #6)
> > Darin, was the dependency on bug #176305 a mistake? It looks unrelated?
> 
> Yes, typo in bug number.

Corrected, it should have been bug 176035.

[Darin Adler]

The scope of this bug is too narrow. I think what we really want to do is to eliminate WebKit project use of C library formatting functions in the printf family. Eliminating String::format is a necessary part of this, but not sufficient to cover the whole task. At some point we might want a bug about that.

[Michael Catanzaro]

(In reply to Darin Adler from comment #9)
> The scope of this bug is too narrow. I think what we really want to do is to
> eliminate WebKit project use of C library formatting functions in the printf
> family. Eliminating String::format is a necessary part of this, but not
> sufficient to cover the whole task. At some point we might want a bug about
> that.

E.g. our logging functions make heavy use of format specifiers. On Unix they boil down to fprintf(stderr, ...) but it takes some investigation every time I need to discover that.

I push a lot of commits changing "%llu" to "%" PRIu64. uint64_t is long long unsigned int on Mac but just long unsigned int on Linux, so using %llu is an error. That's just one example of an annoying difference. Another: Mac has format specifiers that fprintf doesn't accept, e.g. %{private} or something like that. Logging will be a lot more annoying to do without formats, but safer indeed....

[Darin Adler]

After r242014 <https://trac.webkit.org/changeset/242014>, String::format itself is now gone. But as Michael mentions above, logging is still using the former implementation of String::format.

[Darin Adler]

Relanded in <https://trac.webkit.org/changeset/242308>.