| dfaure |
| Frameworks |
We are opening the stream as read only and skipRawData would complain that
it's a write-only device, and skipping needs reading:
QIODevice::skip (QBuffer): WriteOnly device
It was a weird optimization anyway, so just pass the information like the
rest of the data and it works just fine.
BUG: 386364
The check on the bug report passes now, and akregator fetches
all feeds again.
| Automatic diff as part of commit; lint not applicable. |
| Automatic diff as part of commit; unit tests not applicable. |
| src/ioslaves/http/http.cpp | ||
|---|---|---|
| 4910 | I still can't understand why this call present? Doesn't work if it removed? | |
| src/ioslaves/http/http.cpp | ||
|---|---|---|
| 4910 | I understand that the serialised cache tag is part of the command and it's there at the beginning. Skip was so that we wouldn't write on top of it but append after. | |
| src/ioslaves/http/http.cpp | ||
|---|---|---|
| 4910 | Why not just seek ? | |
| src/ioslaves/http/http.cpp | ||
|---|---|---|
| 4910 | Why seek if this is essentially the same? | |
| src/ioslaves/http/http.cpp | ||
|---|---|---|
| 4910 | Current implementation has a 4k buffer on the stack (from Qt source), patched one exclusive copy, seek is faster and cleaner. | |
How does that remove the warning from QIODevice::skip, which we're now calling directly instead of indirectly? I don't get it.
I wonder if the bug is inside QIODevice::skip, which shouldn't CHECK_READABLE in all cases (only when it's going to fall back to reading).
Alternatively we could open the device in ReadWrite mode, no? That might make more sense, for a device that already has some data at the beginning.