IRC Logs

Log of #ghostscript at

 <<<Back 1 day (to 2015/05/03)20150504 
kens chrisl ping08:37.05 
chrisl kens: pong08:43.39 
kens I'm getting nowhere with my gc problem, can you spare some time at some poitn to take a look ?08:43.57 
  Doesn't have to be now, any time you are at a convenient place08:44.12 
chrisl Sure, I'm going to be out for a bit later this morning, so I'll have a look this afternoon - if you can e-mail me details like what test file and command line to use08:44.51 
kens Yes I'll do that now, thanks08:45.01 
chrisl So, just to double check: the top commit is: "8308088e: fix the extern definition of......"08:46.57 
kens Moment08:47.08 
  Yes, that seems to be it (temporary confusion over the clurter run commit)08:47.44 
  Yes that's defintely the one08:48.24 
chrisl Good.....08:48.58 
Robin_Watts chrisl, kens: I looked at gs on the train the other day with a view to this language abstraction thing.10:43.12 
  and in so doing, I think I spotted a bug in gsapi.10:43.22 
kens That's entirely possible, BTW chrisl is away for now, but I'm sure he'll read the lgos10:43.38 
Robin_Watts Anyone care to review:;a=commitdiff;h=b34e6835e6e918dfca8c6558fe00eedcca6660c5 ?10:43.41 
kens One second10:43.49 
Robin_Watts kens: Thanks.10:43.55 
kens Hmm, that's not the way the function is defined in iapi.h is it ?10:45.49 
  Its defined as taking an 'instance' not a ctx10:46.00 
  gsapi_run_string_with_length that is10:46.25 
Robin_Watts kens: iapi.h is bad, it uses lib and instance interchangable.10:46.49 
kens Yes, I see that the actual procedure is defined differently10:47.03 
Robin_Watts We should be passing named types, not void *'s.10:47.10 
kens And does indeed take a ctx10:47.15 
Robin_Watts As it happens... in a moment I will show you my next proposed commit :)10:47.30 
kens So your change makes se4nse to me, it seems to have been wrong before10:47.39 
Robin_Watts kens: Thanks.10:47.50 
  Any thoughts on:;a=commitdiff;h=b5afe3a9594811a6fcc4fde16f36c1f55cab20c710:48.03 
kens So you're basically putting proper type definitions int he functions and declaring an enumerated error ?10:49.42 
Robin_Watts kens: Yes.10:49.52 
kens THose both sound like a good dea to me10:50.07 
Robin_Watts I believe that any calling code that works now, should still work.10:50.20 
  but intent is clearer.10:50.31 
  In an ideal world we'd push the gsapi_error type (or something equivalent) through our entire codebase, so that every int return value that's actually an error was explicit.10:51.17 
  but... ugh.10:51.23 
kens I presume that the reason that void * was used before was to avoid callers code knowing about our data types, that's not going to work now though ?10:51.25 
Robin_Watts kens: callers can know that the thing they are passing is a pointer to an instance.10:52.05 
  But they don't know anything about what the instance is.10:52.15 
kens Sounds good enough10:52.24 
Robin_Watts So we've not made any extra internals public.10:52.34 
  we've just removed the possibility for screwups such as the one I just fixed.10:52.52 
kens OK its not so much internals being public, as tryign to explain to people how to get the structure definitions included :-)10:52.59 
  But yes, that's why ata typing is good10:53.12 
Robin_Watts kens: cool. Will run a sanity cluster test the push the fix. will wait for chrisl to comment on the second one before I push that.10:53.45 
kens You'd probably better pass it via Ray as well, since he's more likely to have dealigns with people who use it10:53.59 
Robin_Watts thanks.10:54.02 
kens Robin_Watts : how do I get text out of a PDF file with mudraw ? (or should I be using mutool ?)10:56.08 
tor8 mudraw -Ftext10:56.41 
  kens: if you're using 1.7 or later10:56.51 
kens I'm using 1.7, which is (obviously) dfifferent to before, hence my confusion :-)10:57.07 
tor8 kens: understandable, I did change the command line quite a bit for 1.710:57.30 
kens Can't I use the -ttt options any more ?10:57.34 
tor8 the -ttt options have been replaced with format selection10:57.43 
  so -F text, stext, or xml10:57.48 
kens aha, OK thanks10:57.54 
tor8 I meant html, not xml10:58.25 
kens And that gives the answers I want for this SO queston as well, thanks again10:58.29 
Robin_Watts Or mudraw -o out.txt in.pdf10:59.53 
sebras tor8: you need to update your batshit-insane scripts now...11:02.11 
tor8 sebras: I've stopped using them in favor of the cluster...11:03.39 
sebras tor8: ah.11:04.04 
tor8 but yes, they would need updating :)11:04.14 
Robin_Watts cluster is running full gs/pcl/xps tests in 12 minutes now. wow.11:08.46 
kens I'd noticed it was pretty quick recently11:09.12 
Robin_Watts tor8: I just got a bug report (on claiming that certain PDFs with annotations + content are only showing annotations in 1.711:09.51 
  I've asked for an example file.11:09.59 
tor8 Robin_Watts: okay. we should be cooking up a new release soon, so do you want to wait a few days for that?11:11.14 
kens Could be the annotatoins are oversized. I had someone send me a file where the post-it annotaton was rendered immense (no /Appearance in the PDF).11:11.16 
Robin_Watts tor8: If it's genuine it sounds bad, so might be worthwhile.11:11.31 
tor8 Robin_Watts: a few commits on tor/master that we missed, adding epub support to the windows viewer11:11.41 
Robin_Watts He's promised me a file lateer today.11:11.43 
sebras tor8: wow... the epub viewer is a bit chatty...11:12.25 
  maybe that's intentional.11:12.50 
tor8 sebras: otherwise you'd think it's crashed or frozen ;)11:12.58 
Robin_Watts tor8: top two look fine.11:13.09 
tor8 Robin_Watts: thanks.11:13.14 
Robin_Watts The others are all in already, right?11:13.31 
tor8 yes.11:13.35 
  top two are the new ones that I forgot to ask you to review last week11:13.46 
Robin_Watts Someone has been burbling about epub stuff on bugzilla - I suspect the answer to the bugs is "yes, we know we don't do that"11:14.09 
tor8 not adding epub to the list of files we support on the win32 viewer is embarrassing, but only goes to show how much I use windows :)11:14.20 
sebras tor8: hm... #cfs_0 { font-size: 90% } why would mupdf complain about "css syntax error: invalid color" on this one?!11:15.36 
tor8 #11:15.48 
sebras tor8: oh, so you have to have the element name there for it to be a valid selector?11:19.33 
  just having the id there is not sufficient..?11:19.45 
tor8 it might be a genuine slip-up on my part, not lexing #{name} properly11:20.10 
  and always assuming #{name} is a color11:20.23 
sebras I guess the intent is for the block to apply to <anything id="cfs_0">...</anything>11:21.08 
  but I'm having a hard time verifying this in the css spec.11:21.25 
  tor8: this is the best I can do. last senten of second paragraph.11:22.47 
  tor8: though, maybe you have to write *#cfs_0..?11:23.47 
tor8 sebras: yeah, it's a thinko in my css lexer11:39.55 
  the parser already looks at '#' as the selector11:40.05 
  but that code path has apparently never been tested >.<11:40.12 
sebras tor8: good thing I tested with my .epub here!11:55.28 
kens chrisl I made a bit more progress wiht my problem. I hope you saw my email ?13:20.51 
chrisl kens: I did - tbh, it confused me even more.....13:21.09 
kens Ah, I thnk I more or less understand what's going on. I'll opena private chat, no reason to annoy everyone else13:21.30 
tor8 Robin_Watts: fixes for the two epub bugs on tor/master14:31.10 
  bah, don't review yet, I just spotted an error14:34.12 
Robin_Watts ok. Clearly I fail as a reviewer, cos I didn't spot it.14:36.40 
tor8 '#000000' will fail to parse as a color14:48.30 
  it'll get '#' and then number (not keyword)14:48.39 
  I only tested on '#aa0000' when bashing out the changes14:49.00 
  Robin_Watts: okay, updated commits on tor/master14:59.53 
Robin_Watts tor8: is rgb(10 10 10) valid? or just rgb(10,10,10) ?15:31.32 
tor8 just rgb(10,10,10)15:33.12 
Robin_Watts tor8: OK, all look plausible to me.15:36.28 
tor8 Robin_Watts: thanks.15:45.33 
kens TIMe to go, goodnight all16:20.46 
UukGoblin <- here are the messages leading to a segfault when trying to convert to jpeg files16:36.27 
  file displays OK in xpdf (although slowly and with lots of warnings)16:36.46 
  also, pdftk 2.02 is able to convert it fine, however 1.44 had problems16:37.02 
chrisl UukGoblin: I would first suggest trying a newer Ghostscript version - we just released 9.16 recently16:38.06 
UukGoblin chrisl, oh, I've just updated to debian latest :-S OK, let me try compile that16:38.31 
  chrisl, but, could you do me a favour and try to convert that PDF I linked if you have 9.16 on your box?16:39.00 
  oh, there's a binary, ok16:40.27 
chrisl You've also missed off the output file which won't help....16:40.46 
UukGoblin w00t, all fixed, works like a charm in 9.16, sorry for the noise :-)16:42.04 
  output file is there, 2nd/3rd args16:42.26 
chrisl NP16:42.26 
  Oh, yeh, it threw an error for me because the directory doesn']t exist here16:43.16 
UukGoblin ah, right, sorry about that16:51.02 
 Forward 1 day (to 2015/05/05)>>>