Log of #ghostscript at irc.freenode.net.

Search:
 <<<Back 1 day (to 2019/02/11)20190212 
sebras chrisl: may I trouble you with two reviews?17:03.43 
  chrisl: http://git.ghostscript.com/?p=user/sebras/ghostpdl.git;a=commitdiff;h=be1a327733c10aafa99f7772b0f3aa287df58a1f which is a precondition for 17:03.51 
  chrisl: this one http://git.ghostscript.com/?p=user/sebras/ghostpdl.git;a=commitdiff;h=2486669a97238eef7a23dcbe4d68ef733820170a17:04.01 
chrisl sebras: give me a few moments to finish a commit message17:04.32 
sebras the aim is to have the jbig2 decoders use a custom allocator based on gs'17:04.33 
  chrisl: no worries.17:04.52 
chrisl sebras: So..... Is there a specific reason you opted to have jbig2_global_ctx_free() return the allocator, rather than just have it incumbent on the caller to store it's allocator? (I'm trying to remember how the other shared libs do this)17:14.34 
sebras if I did it that way s_jbig2decode_release() can get the allocator back without having it stored elsewhere.17:15.41 
  besides the mem pointer the allocator can be constant really.17:16.16 
  so another option would be to change jbig2dec so that jbig2_ctx_new() takes a separate allocator_data and stores it on its own.17:16.44 
chrisl My only worry is the change to the ABI might irritate some folks17:16.53 
sebras a problem in s_jbig2decode_make_global_data() is that there is really nowhere to store it.17:17.49 
  perhaps to return something other than Jbig2GlobalCtx in *result17:18.12 
  basically wrap Jbig2GlobalCtx and Jbig2Allocator in a new struct and have s_jbig2decode_make_global_data() and s_jbig2decode_free_global_data() cast to/from it upon alloc/release.17:19.28 
chrisl TBH, I like the way you have it, I'm just wondering about the hassle we'll get for changing the interface again17:20.17 
sebras I like this way because it is clear at what point the library will stop using the allocator.17:20.46 
  I'll whip up a patch with the alternative just to see if it is feasible.17:21.19 
  chrisl: I remember you being wary of the API/ABI last time and then I changed it nonetheless and then people complained...17:21.41 
  ergh... 6 differences in the cluster. I'm bmpcmping and hoping that they amount to nothing.17:23.18 
  chrisl: btw, I ran the current version with the gpdlsanitize target on a pdf that I know has jbig2 images in it and it works fine now.17:25.01 
chrisl sebras: I think we should go with what you have, but if you can add a comment to CHANGES - even if it's just something to remind me to write something more detailed.17:26.02 
sebras chrisl: so you advice against the wrapping/unwrapping of the Jbig2Allocator and global context?17:26.48 
chrisl sebras: I think that complicates things more than I like17:27.12 
sebras alright.17:27.24 
  I'll add something to CHANGES under version 0.16 in that case.17:27.54 
  chrisl: thanks for reviewing btw.17:28.08 
chrisl NP17:28.20 
sebras chrisl: yey, the cluster and bmpcmp went through without issues.17:44.13 
  chrisl: then I'll add something to CHANGES, spell check it and merge..?17:44.31 
chrisl sebras: Sounds good17:44.40 
sebras sebras: since the changed functions used to return void I don't forsee that this would be a major issue.17:45.32 
  chrisl: btw, I'll just leave the date for 0.16 empty.17:46.06 
  there might be more changes...17:46.09 
chrisl Sure - as I said, I'll probably add extra words for the release. I just need you to add enough in there to remind me. Otherwise I'll end up writing "Stuff changed....."17:47.18 
sebras "Bug fix release." seems to be popular... :)17:47.39 
chrisl Yeh, but if we change the interface, I feel we need to be more explicit.17:48.16 
sebras I think we did last time, but we weren't. or I wasn't and I that certainly didn't help you.17:49.29 
  chrisl: how does the change in c972f9f3a look?17:50.12 
  s/change/CHANGES/17:50.22 
chrisl Can I suggest adding "API change" or something? Again, mainly as a reminder17:51.35 
sebras as far as I can tell this change will be backwards compatible.17:57.13 
  do you agree?17:57.16 
chrisl I suspect it'll depend on compiler options18:00.25 
  But yeh, normally it should only introduce a warning18:01.04 
sebras chrisl: alright, "API change" has been added. 18:02.12 
  chrisl: I'll wait for you to update to 0.16 at release time, I hope that's ok?18:02.47 
chrisl sebras: Yeh, as I said, I just need a reminder that more was required than "bug fix release" :-)18:04.54 
sebras ok, then I'll merge to master now. yey! :)18:05.16 
 Forward 1 day (to 2019/02/13)>>> 
ghostscript.com #mupdf
Search: