On 18.03.2018 16:04, Daniel Shahaf wrote: > Branko Čibej wrote on Sat, 17 Mar 2018 20:27 +0100: >> On 17.03.2018 20:24, br...@apache.org wrote: >>> >>> + * r1825979 >>> + Minor clarification to docstring. >>> + Justification: >>> + Trivial documentation fix. >>> + Votes: >>> + +1: danielsh, jamessan >>> + -1: brane (the docstring "clarification" is wrong, since >>> + we do not propagate the compression level to LZ4) >> >> It seems I failed to this change on the list when it was originally made >> on trunk. I believe r1825979 should be reverted because the change is >> not correct. > Thanks for the review. Before r1825979 the docstring implied that wire > compression always used zlib; r1825979 corrects that misimplication. > > I grant that it's not clear from the docstring that compression levels > aren't honoured by all compression functions in use. It would be good to > document that, possibly by reference (to svn_delta.h, I guess?). I'm afraid > I don't have a specific suggestion at the moment; does anyone else?
How about this: Index: svn_ra_svn.h =================================================================== --- svn_ra_svn.h (revision 1827142) +++ svn_ra_svn.h (working copy) @@ -179,7 +179,9 @@ typedef svn_error_t *(*svn_ra_svn_edit_callback)(v * * Either @a sock or @a in_stream/@a out_stream must be set, not both. * @a compression_level specifies the desired network data compression - * level (zlib/lz4) from 0 (no compression) to 9 (best but slowest). + * level from 0 (no compression) to 9 (best but slowest). The effect + * of the parameter depends on the compression algorithm; for example, + * it is used verbatim by zlib/deflate but ignored by LZ4. * * If @a zero_copy_limit is not 0, cached file contents smaller than the * given limit may be sent directly to the network socket. Otherwise, -- Brane
