Bug 59190 – Support for alignment for the Sun/Oracle Studio compilers

• Status: UNCONFIRMED
• Component: Web Template Framework

[Ben Taylor]

The check for alignment that was in wtf/Vector.h got moved to Alignment.h, but
the COMPILER(SUNCC) didn't go with it.  Studio 12.1 supports alignment macros.
Studio 12 does not, but we can compensate by overcommiting the buffer by 64 bytes
and then aligning the pointer manually inside the overcommited buffer.

[Ben Taylor]

Created attachment 91160 [details]
patch to support alignment macros for SunStudio >= 12.1, overcommit buffer by 64 bytes for Studio 12 and manually alighn buffer

This was fixed earlier in Vector.h, but was missed in the transition to Alignment.h. Further research
showed Studio 12 ignored the alignment macro, so added code to do manual alignment just for Solaris with SS12.

[WebKit Review Bot]

Attachment 91160 [details] did not pass style-queue:

Failed to run "['Tools/Scripts/check-webkit-style', '--diff-files', u'Source/JavaScriptCore/ChangeLog', u'Source..." exit_code: 1

Source/JavaScriptCore/wtf/Vector.h:61:  This { should be at the end of the previous line  [whitespace/braces] [4]
Total errors found: 1 in 3 files


If any of these errors are false positives, please file a bug against check-webkit-style.

[Ben Taylor]

Created attachment 91170 [details]
Updated patch to fix the check-webkit-style error.

Same comments as previous.

[Eric Seidel (no email)]

Comment on attachment 91170 [details]
Updated patch to fix the check-webkit-style error.

View in context: https://bugs.webkit.org/attachment.cgi?id=91170&action=review

> Source/JavaScriptCore/wtf/Vector.h:62
> +        AlignedBufferChar *buffer()

Style.  * always pins to the type, not the name in WebKit.

> Source/JavaScriptCore/wtf/Vector.h:64
> +            AlignedBufferChar *ptr = oversizebuffer;

style again.

> Source/JavaScriptCore/wtf/Vector.h:66
> +            ptr += 64 - (reinterpret_cast<size_t>(ptr) & 0x3f);
> +            return ptr;

I'm not sure why this is written out in 3 lines, seems to be a single-line worth of code.  I don't think the locals help readability here.