Having deque of pointers to objects of about uint32
was not really efficient, plus we had the "Weak-vtable"
warning.

keep everything in a union, manually resolve the right
function to call. Now we have a deque of operations.
Only block and reorder need further allocations.

Signed-off-by: Luca Fulchir <luker@fenrirproject.org>

Only naively ported. We now have a full vtable for each
object.

TODO: eliminate the vtable, check if size() is really used
(hint: don't think so).

Signed-off-by: Luca Fulchir <luker@fenrirproject.org>

Signed-off-by: Luca Fulchir <luker@fenrirproject.org>

While I was at it, Octet got its own file.

Signed-off-by: Luca Fulchir <luker@fenrirproject.org>

Signed-off-by: Luca Fulchir <luker@fenrirproject.org>

Work towards #16



Signed-off-by: Luca Fulchir <luker@fenrirproject.org>

Signed-off-by: Luca Fulchir <luker@fenrirproject.org>

Signed-off-by: Luca Fulchir <luker@fenrirproject.org>

Technically, we broke the old C RAW API.
but it was just created, and we are in prealpha.
And we only put things const. so whatever.

Signed-off-by: Luca Fulchir <luker@fenrirproject.org>

Compared to v0.1.X we already broke the whole API,
and created 3 (three) new ones (C++ Raw, C RAW, C RFC),

Since our versioning scheme tracks the API incompatibilities
as the first version number, this seems like a good moment
to bump it up.

Things are finally stabilizing, too.
Of course, it's still a prealpha....
(read: precomputation sharing is incomplete/not tested enough)

Signed-off-by: Luca Fulchir <luker@fenrirproject.org>

Better command line parsing, added block listing.
working for #9

Signed-off-by: Luca Fulchir <luker@fenrirproject.org>

Minor API changes. everything is const now,
same as in the C RFC API

Signed-off-by: Luca Fulchir <luker@fenrirproject.org>

Thank you GCC, we had a bug.

Signed-off-by: Luca Fulchir <luker@fenrirproject.org>

Signed-off-by: Luca Fulchir <luker@fenrirproject.org>

Close #11. Seems good enough, but let's not call it
final yet.

Minor tweaks to the CPP API, too

Signed-off-by: Luca Fulchir <luker@fenrirproject.org>

...although you will rarely need it for the encoder.

Signed-off-by: Luca Fulchir <luker@fenrirproject.org>

should not be necessary, we made things typesafe.
But there are strange people around...

Signed-off-by: Luca Fulchir <luker@fenrirproject.org>

And finally fix #3



Testing is needed, though.

Signed-off-by: Luca Fulchir <luker@fenrirproject.org>

Spaces, not tabs.

Actually, I liked tabs much more. But it seems a lot
of IDEs and general reader do not handle them correctly,
or do not let you configure "1 tab == 4 spaces", so the code
was not always readable.

"git diff -w" for this commit is en empty list.

Signed-off-by: Luca Fulchir <luker@fenrirproject.org>

This was originally meant only to eradicate the interleaving
code, so that the raw API could be slightly faster.

Turns out a lot had to be done for that, and the API changed
to reflect those changes.

The RAW_Encoder template now has enabled/disabled methods,
so that you can only use the right one, depending on
whether you choose interleaving or not.

While doing all of this, I realized a nice way to let the user
choose the correct symbol number for the block. +1 for typesafety.

And since the commit became really big, I threw in a couple
of useless style changes, bringing back everything to 80 columns.

Really, I should be doing smaller commits.

Signed-off-by: Luca Fulchir <luker@fenrirproject.org>

Ticket #3 not yet finished, blocked by #7

Signed-off-by: Luca Fulchir <luker@fenrirproject.org>

An other commit towards #2

Signed-off-by: Luca Fulchir <luker@fenrirproject.org>

bugfix & do not use LZ4 for now.
not really supported by the cache.

Signed-off-by: Luca Fulchir <luker@fenrirproject.org>

Signed-off-by: Luca Fulchir <luker@fenrirproject.org>

use "using" instead of duplicating code.

Signed-off-by: Luca Fulchir <luker@fenrirproject.org>

If we get told that there is no more input for a block
we can return error and avoid waiting indefinitely.

Signed-off-by: Luca Fulchir <luker@fenrirproject.org>

Still hunting an other one.
Many thanks to the developers of rr.

Signed-off-by: Luca Fulchir <luker@fenrirproject.org>

Was not singleton, and some loking problems.
doesn't seem to hang up anymore

Signed-off-by: Luca Fulchir <luker@fenrirproject.org>

Start portability for VC++14. Still unfinished.

Also simplify a bit locking in RFC calls.

Signed-off-by: Luca Fulchir <luker@fenrirproject.org>

Making the raw API actually try to decode
might be a good thing...

Also resolved race conditions in CLI tool

Signed-off-by: Luca Fulchir <luker@fenrirproject.org>

Also: modify a bit the API for 32 bits.
The user should not expect RQ to manage more memory
than the 32 bit machine can handle. This means that
some methods should work with and return size_t, not uint64_t

Signed-off-by: Luca Fulchir <luker@fenrirproject.org>

Signed-off-by: Luca Fulchir <luker@fenrirproject.org>

Not using LZ4 might cause function to reach
end without properly returning.

Signed-off-by: Luca Fulchir <luker@fenrirproject.org>

Modified the RAW API (again) to make it more usable.
"bytes" parameter ended up being confusing.

Cli tool should nw be able to ancode/decode properly

Signed-off-by: Luca Fulchir <luker@fenrirproject.org>

Decoder still does not work.

Signed-off-by: Luca Fulchir <luker@fenrirproject.org>

See #9. Uses new RAW API.
Completely untested. At least it compiles.

Also, did I forget to push Bitmask.hpp?
Awkward...

Signed-off-by: Luca Fulchir <luker@fenrirproject.org>

Dropping symbols and retrying does not make that much sense.
Just tracking whether we can retry or not is enough.

Signed-off-by: Luca Fulchir <luker@fenrirproject.org>

Signed-off-by: Luca Fulchir <luker@fenrirproject.org>

Other symbols should not be exported, but are needed right now.

Signed-off-by: Luca Fulchir <luker@fenrirproject.org>

This way we don't need to sync on which algorithm
to use, as that would introduce locking.

Signed-off-by: Luca Fulchir <luker@fenrirproject.org>