On 29 July 2013 13:20, <marti.ma...@littlecms.com> wrote: > I'm taking a look on this. The cat's state may be caused by > a serialization feature: a profile that has never been flushed > to disk or memory, have a representation of internal data > which may be different. Let me a day or two to do the > necessary testing.
Great, thanks. A little more info: The bugs only seems to be triggered when you do cmsSaveProfileToMem() to get the profile size, then cmsSaveProfileToMem() again to populate the data blob. If you allocate a 16Mb lump of memory, and then do cmsSaveProfileToMem() just once, the bug doesn't seem to trigger. To add to your worries, I tried to work around it by allocating a lump of memory like above, and then calling cmsSaveProfileToMem() just once. This writes the profile, but does not set BytesNeeded, so we don't know how many bytes to write when saving the file. In colord to get things working I've used cmsSaveProfileToFile and then reading back the temp file. This sucks as this is much slower (although writing profiles isn't in the hot path by any means) and I can't just use cmsSaveProfileToFile() in libcolord as it writes using GFile, which allows me to save files on remote shares and that kind of thing. This is the colord workaround I've used: https://github.com/hughsie/colord/commit/0baa06134a86b493a07156531ade5fc15b26618c -- better ideas welcome. Richard. ------------------------------------------------------------------------------ See everything from the browser to the database with AppDynamics Get end-to-end visibility with application monitoring from AppDynamics Isolate bottlenecks and diagnose root cause in seconds. Start your free trial of AppDynamics Pro today! http://pubads.g.doubleclick.net/gampad/clk?id=48808831&iu=/4140/ostg.clktrk _______________________________________________ Lcms-user mailing list Lcms-user@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/lcms-user
