RESOLVED FIXED 109092
IndexedDB: Implement SharedBuffer version of put()
https://bugs.webkit.org/show_bug.cgi?id=109092
Summary IndexedDB: Implement SharedBuffer version of put()
Alec Flett
Reported 2013-02-06 14:00:04 PST
IndexedDB: Implement SharedBuffer version of put()
Attachments
Patch (23.09 KB, patch)
2013-02-06 14:39 PST, Alec Flett
no flags
Patch for landing (23.53 KB, patch)
2013-02-15 12:35 PST, Alec Flett
no flags
Alec Flett
Comment 1 2013-02-06 14:39:50 PST
Alec Flett
Comment 2 2013-02-06 14:40:59 PST
jsbell/dgrogan - here's the first cut at using SharedBuffer abarth - care to review - not just the WebKit API but my uses in core IDB?
Alec Flett
Comment 3 2013-02-06 14:50:20 PST
Also I fully expect this patch to fail dramatically (not sure why cr-linux-debug even passed) until the chromium side (https://codereview.chromium.org/12217049/) lands.
Adam Barth
Comment 4 2013-02-06 15:51:37 PST
Comment on attachment 186923 [details] Patch View in context: https://bugs.webkit.org/attachment.cgi?id=186923&action=review > Source/WebCore/Modules/indexeddb/IDBObjectStore.cpp:234 > + Vector<char>* valueBytesSigned = reinterpret_cast<Vector<char>*>(&valueBytes); Is this reinterpret_cast valid? It seems slightly scary.
Adam Barth
Comment 5 2013-02-06 15:52:14 PST
Comment on attachment 186923 [details] Patch This seems fine, modulo the question about reinterpret_cast above.
Alec Flett
Comment 6 2013-02-06 15:53:47 PST
Comment on attachment 186923 [details] Patch View in context: https://bugs.webkit.org/attachment.cgi?id=186923&action=review >> Source/WebCore/Modules/indexeddb/IDBObjectStore.cpp:234 >> + Vector<char>* valueBytesSigned = reinterpret_cast<Vector<char>*>(&valueBytes); > > Is this reinterpret_cast valid? It seems slightly scary. It's totally scary. It was the only way I could find to adopt a Vector<uint8_t> into SharedBuffer.
Adam Barth
Comment 7 2013-02-06 16:01:45 PST
> It's totally scary. It was the only way I could find to adopt a Vector<uint8_t> into SharedBuffer. On webkit-dev you talked about changing Vector<uint8_t> to Vector<char>. Folks seemed unsure about whether that was a good idea because the reason was to play nice with LevelDB. Perhaps playing nice with SharedBuffer is a reason that tips the scales in favor of switching to Vector<char>.
Alec Flett
Comment 8 2013-02-15 12:35:35 PST
Created attachment 188620 [details] Patch for landing
WebKit Review Bot
Comment 9 2013-02-15 13:03:29 PST
Comment on attachment 188620 [details] Patch for landing Clearing flags on attachment: 188620 Committed r143037: <http://trac.webkit.org/changeset/143037>
WebKit Review Bot
Comment 10 2013-02-15 13:03:34 PST
All reviewed patches have been landed. Closing bug.
Note You need to log in before you can comment on or make changes to this bug.