any idea what buff_src.st_size < 0x7FFFFFFF was for?

reduce the overhead of processedSize dbus progress notification in each iteration since it is called less often, adding some more speed.

From what I recall in copy job the processed size in emitted by a timer and not every time a chunk got processed

In D28555#641276, @dfaure wrote:

any idea what buff_src.st_size < 0x7FFFFFFF was for?

None, it does not makes sense to me, perhaps, we had an implementation limitation.

In D28555#641277, @broulik wrote:

reduce the overhead of processedSize dbus progress notification in each iteration since it is called less often, adding some more speed.

From what I recall in copy job the processed size in emitted by a timer and not every time a chunk got processed

You are right, it is emitted every 100ms in SlaveBase::processedSize, so this part was overzealot.

@dfaure: probably this:

man sendfile, NOTES

The  original Linux sendfile() system call was not designed to handle large file offsets.  Consequently, Linux 2.4 added sendfile64(), with a wider type for the offset argument.  The glibc sendfile() wrapper function transparently deals with the kernel differences.

Sounds like the size check should stay then.

In D28555#641390, @dfaure wrote:

Sounds like the size check should stay then.

Pessimizing sendfile for everyone, for the sake of Linux 2.2.x users?