I attach a new patch trying to adapt the LZW filter implementation to the new filter API.
Applied. 2. Because of the cache size stuff, those embarrassing weird goto's are kept. Maybe we can consider allowing the filters specify a "minimum-cache" parameter to avoid non-necessary code bloat and/or efficiency tradeoffs? It is indeed something to consider: it is quite unlikely for applications to use cache sizes like 1 or 10. It is good to not require a minimum cache size, as long as it is possible, but a reasonable minimum cache size due to efficiency issues would be considered. 4. I factored out the test cases sent by Brad Hards using fixtures to allocate/deallocate the stream. This may cause an overhead in the overall runing time as the fixture is setup/teareddown on every test cases. I still think that it would be useful to factor-out other tests to use this fixture and so avoid setup/dealloc boilerplate -I did not do it in case their authors disagree. I agree with that approach. Don't worry about the running time of the tests. 5. I disabled the tokenizer utility, it seems that a header file was forgotten in the last commit. Yep, utils/pdf-tokeniser.h was missing. I incorporated it. -- Jose E. Marchesi <[email protected]> http://www.jemarch.net GNU Project http://www.gnu.org
