| <<<Back 1 day (to 2013/09/16) | 2013/09/17 |
Robin_Watts | kens, chrisl: ping | 12:48.22 |
| Type 3 Images in postscript are masked ones, right? | 12:48.33 |
kens | chrisl is away but I can pong :-) | 12:48.35 |
| type 3 images are a kind of mask | 12:48.44 |
| they are an image and a specific mask | 12:49.16 |
Robin_Watts | right. | 12:49.30 |
kens | type 4 are chroma keyed masked images | 12:49.33 |
Robin_Watts | so given we handle masks and images as 2 separate things... | 12:49.50 |
| and that we never interpolate the mask... | 12:49.57 |
kens | THen we probably get stupid resutls yes | 12:50.07 |
Robin_Watts | does it seem reasonable to you that we should disable interpolation on the image part too? | 12:50.12 |
kens | TO me, yes it certainly does | 12:50.24 |
| interpolation is 'device dependent' anyway | 12:50.34 |
Robin_Watts | Cool. 1 line fix then :) | 12:51.45 |
kens | section 4.10.5: | 12:51.49 |
| Image interpolation may not always be performed for | 12:51.49 |
| some classes of image or on some output devices. | 12:51.49 |
| There's a note in section 4.10.6 that type 3 images are often simulated using and image and a clip in level 2 PostScript. which wuold have the effect of interpolating teh image, but not the mask, just as we currently do | 12:53.06 |
| You could ask ray to try it on a CPSI device | 12:53.51 |
Robin_Watts | kens: This is a PDF input file. | 12:54.18 |
kens | Ray'#s CPSI doesn't handle PDF input ? | 12:54.44 |
Robin_Watts | I thought CPSI was PS only. | 12:54.55 |
kens | It's posible to haul the image out and put it in a PostScript file | 12:55.01 |
| Later versions of CPSI handle PDF too | 12:55.09 |
Robin_Watts | runs clusterpush. Can't see it doing much, but... | 12:56.46 |
kens | Interesting, there seems to be no way to reset the CTM in PDF | 13:05.32 |
| Other than q/Q | 13:05.41 |
Robin_Watts | or putting the inverse matrix into a cm. | 13:28.48 |
kens | Hmm, yes I guess that would work | 13:29.15 |
Robin_Watts | Though q/Q is better to avoid rounding errors. | 13:30.19 |
kens | can't do it | 13:30.30 |
| I have no idea what save level we are at at the time | 13:30.46 |
paulgardiner | Hi tor7. There's a few iOS commits on paulg/master when you have a moment. | 13:31.14 |
tor7 | paulgardiner: there's a typo in the second commit message | 13:44.04 |
kens | Noldorin what was your question about arrays last night ? | 13:44.19 |
Noldorin | kens, well i'm trying to pass an array into a script by the command line and then loop through it⦠wondering how to do that | 13:44.48 |
kens | 'pass an array into a script' ? not sure wht you mean there | 13:45.12 |
| I you want to declare an array on the command line then as Ray said you need to use the -c switch to write some PostScript | 13:45.46 |
| In PostScript, you can use the 'forall' operator to access each element of an array in turn. See the PLRM for details | 13:46.14 |
tor7 | paulgardiner: I'm not sure if I agree with having the one file per utility function in the ios app. having the classes split out makes sense, but I'd make for example fitPageToScreen as a static function near the place it is used | 13:46.20 |
paulgardiner | tor7: ah the 'g'. I stared at it for ages before I could see it | 13:46.33 |
tor7 | but if you feel this way is better, I won't argue | 13:46.35 |
paulgardiner | I've put most of the static functions in with the classes, but fitPageToScreen was used by two classes | 13:47.19 |
tor7 | oh, right. "common.m" perhaps? | 13:47.37 |
| (or just feel free to ignore me on this complaint) | 13:48.05 |
| the three patches look good to me, if you fix the typo (or better yet, reword to a more active voice "use the background thread in CGDataProvider's callback") | 13:49.08 |
paulgardiner | Same worries occured to me, I must admit. I kept search separate because I thought it might be worth making into a class, and reaturn an NSArray or the like | 13:49.39 |
| Do you think fitpage, search and state would be best combined into common? I have no strong oppinion. | 13:50.58 |
tor7 | paulgardiner: the search_page function should really be a common wrapper function for all clients to use | 13:50.58 |
| having to set up and tear down the device is IMO too low level for most customers | 13:51.13 |
| I think it'd be better if the non-class-specific utility functions lived in one file given how small they are currently | 13:51.48 |
paulgardiner | You mean search_page should be internalised to the library? | 13:51.52 |
tor7 | paulgardiner: yes. I have a start on a set of "wrapper" functions like that one sitting on a branch somewhere. | 13:52.13 |
| something I toyed with on the flight | 13:52.20 |
paulgardiner | Oh okay. I'll collect the bits together for now. | 13:53.25 |
kens | chrisl ping | 14:24.35 |
chrisl | kens: pong | 14:24.46 |
kens | I added htis to int.mak: | 14:24.53 |
| # ------------------ Support high level Forms ------------------ # | 14:24.54 |
| form_=$(GLOBJ)zform.$(OBJ) | 14:24.54 |
| $(GLD)form.dev : $(LIB_MAK) $(ECHOGS_XE) $(form_) | 14:24.54 |
| $(SETMOD) $(PSD)form $(form_) | 14:24.54 |
| $(ADDMOD) $(PSD)form -oper zform | 14:24.54 |
| $(PSOBJ)zform.$(OBJ) : $(PSSRC)zform.c $(OP) | 14:24.54 |
| $(PSCC) $(PSO_)zform.$(OBJ) $(C_) $(PSSRC)zform.c | 14:24.55 |
| And I'mgetting a buiold fail: | 14:25.15 |
| psi/int.mak:1735: *** missing separator. Stop. | 14:25.15 |
| ../gs/psi/int.mak:1735: *** missing separator. Stop. | 14:25.15 |
| make: *** [xps-clean] Error 2 | 14:25.15 |
| What did I do wrong ? | 14:25.25 |
chrisl | I'd guess it's spaces rather than tab between the target and deps | 14:25.52 |
kens | Oh, needs to be tabs ? Didn't know that | 14:26.13 |
| thanks | 14:26.16 |
chrisl | Well, let's if I'm right | 14:26.38 |
kens | so tabe before and after the ':' separators ? | 14:27.14 |
chrisl | Just after: target:<tab>deps | 14:27.46 |
kens | OK did it both before and after, lets see what happens :-) | 14:28.07 |
chrisl | Erm, I think..... | 14:28.10 |
kens | Looks like the $(PSCC) should start with a tab | 14:29.57 |
| and VS keeps replacing it with spaces..... | 14:30.18 |
chrisl | I never use VS for makefiles | 14:31.11 |
kens | Well I was editing everything else there.... | 14:31.27 |
| I cheated and cut/paste from another line | 14:31.39 |
chrisl | Yeh, it's just with VS setup for "correct" source formatting, it makes it wrong for makefiles. | 14:32.16 |
kens | Indeed | 14:32.23 |
| OK I *htiunk* the white space is right now | 14:32.54 |
Robin_Watts | I could remove the password from the regression dashboard now. Anyone have any opinion on whether I should? | 14:36.16 |
chrisl | Robin_Watts: I'd just leave it, personally | 14:42.54 |
kens | Hmmwill ahve to do a commit, because I added files. Branch time :-) | 14:52.26 |
henrys | my meeting plan today was to review the high priority projects in the agenda - we didn't get to it at the meeting we can do that all together at the gs meeting or separately if mupdf'ers can't stick around. | 14:55.15 |
Robin_Watts | Either is fine with me. | 14:56.11 |
AmR|EiSa | Hello | 14:58.06 |
Robin_Watts | We haven't done "someday projects" and everything after it, right? | 14:58.25 |
henrys | Robin_Watts: actually I don't recall where we stopped but we were not finished. I have to keep the high priority projects up to date for miles⦠don't kill the messenger ;-) | 14:59.48 |
Robin_Watts | henrys: I wasn't disagreeing that it was a sensible thing. | 15:00.09 |
mvrhel_laptop | henrys: that works well for me today. I have to run my daughter to school at 8:15. Will be back by 8:45 | 15:00.11 |
| Robin_Watts: I see that you "fixed" the bug that I gave you | 15:00.58 |
Robin_Watts | henrys: For me, progressive loading is "done". So unless I hear different, my high priority thing is "jni bindings" I guess. | 15:01.04 |
| mvrhel_laptop: Yeah. You gonna disagree with my fix? | 15:01.16 |
mvrhel_laptop | Robin_Watts: no. it was just not what I had in mind at the time. | 15:02.14 |
henrys | Robin_Watts: we do have a bit of customer pressure for svg output, inclined to give you those 2. | 15:02.37 |
Robin_Watts | henrys: SVG output has just about got to the point where fonts are the next big hurdle. | 15:03.11 |
AmR|EiSa | I want use gs907w32.exe on my installer, how I can run gs907w32.exe auto ? | 15:03.32 |
Robin_Watts | so... I'll need help from tor, probably. | 15:03.57 |
henrys | Robin_Watts: I wonder if we should poke Raed and give him what we have and see what he complains about. | 15:04.45 |
Robin_Watts | henrys: well, he'll complain about fonts, cos currently every font is "non-embedded Helvetica" | 15:05.12 |
mvrhel_laptop | :) | 15:05.21 |
chrisl | AmR|EiSa: assuming you are fully compliant with the licensing, see the fifth bullet point: http://ghostscript.com/FAQ.html | 15:05.23 |
henrys | Robin_Watts: oh I assumed you were creating outlines. | 15:05.49 |
Robin_Watts | henrys: No. MuPDF doesn't do 'decomposition' of device calls. | 15:06.21 |
henrys | Robin_Watts: can you do images for fonts? | 15:07.06 |
Robin_Watts | (gs devices have defaults that call back into the device to decompose high level calls to simpler and simpler things. MuPDF doesn't.) | 15:07.11 |
| henrys: I could do images or paths. | 15:07.44 |
| but it would me refactoring the existing code. | 15:07.55 |
AmR|EiSa | Robin_Watts: done, Thanks :). | 15:07.56 |
Robin_Watts | Currently the code to call freetype lives in the draw device and only there. | 15:08.49 |
| I'd need to try to extract it out a bit so it could be called from other devices too. | 15:09.09 |
| but I can look at that. | 15:09.14 |
| BUT... doing fonts properly is a better idea, I suspect, and may not be much more work. | 15:09.39 |
paulgardiner | henrys: I guess my main high-priority task is updating the iOS app. Is that right? | 15:09.50 |
henrys | Robin_Watts:yes but images and paths are going to be necessary in some fallback situations anyway. | 15:11.05 |
Robin_Watts | henrys: are they? | 15:11.13 |
kens | Only if you go to lower levesl of PDF | 15:11.46 |
| higher to lower that is | 15:11.54 |
Robin_Watts | kens: This is svg out. | 15:12.08 |
kens | Oh then you will need to render shadings | 15:12.18 |
| and most stransparency | 15:12.28 |
Robin_Watts | kens: we are discussing fonts. | 15:12.37 |
tor7 | I have "decomposition device" sitting here in a "future projects" text file :) | 15:12.47 |
kens | I htought henrys was being more general | 15:12.54 |
| But I doubt svg supports type 3 or type 1 fonts | 15:13.23 |
Robin_Watts | type3 fonts would need to fallback to streams of operators, yes. | 15:13.51 |
kens | TT/CFF/OTF should be ok | 15:13.52 |
Robin_Watts | type1 fonts can be converted to other fonts. | 15:14.00 |
henrys | yes I was just speaking of fonts⦠| 15:14.04 |
| type42? | 15:14.13 |
kens | type42 == TT | 15:14.21 |
| type 1 convert to CFF and then make an OTF font | 15:14.30 |
chrisl | PDF doesn't do Type 42 | 15:14.31 |
kens | THat too | 15:14.43 |
henrys | I'm just assuming SVG does not come near PDF font support but I might be mistaken. | 15:15.03 |
Robin_Watts | Right, so, stuff breaks down into 'type3' and 'non-type3' cases, I think. | 15:15.12 |
kens | with the | 15:15.21 |
Robin_Watts | type3 needs to be handled by outputting streams of operators. | 15:15.25 |
kens | caveat of conversion for type 1 | 15:15.26 |
Robin_Watts | and non-type3 either needs to go to outputting fonts (with conversion as required) or to fallback to images/paths. | 15:15.58 |
mvrhel_laptop | svg has transparency why do you need to have it rendered kens? if the blending is normal at least | 15:16.30 |
henrys | chrisl: I meant truetype. | 15:16.32 |
kens | mvrhel_laptop : for the non-normal cases | 15:16.41 |
mvrhel_laptop | and it has gradiants too | 15:16.51 |
henrys | Robin_Watts: how would you handle any incremental font? | 15:17.12 |
kens | mvrhel_laptop : mesh shadings ? Gouraud ? | 15:17.12 |
| nehrys no incremental fonts in PDF | 15:17.24 |
Robin_Watts | henrys: a what now? | 15:19.15 |
henrys | kens:you do not have to embed an entire font in pdf to use it. | 15:19.17 |
Robin_Watts | henrys: You mean a subset font? | 15:19.28 |
henrys | Robin_Watts: yes | 15:19.39 |
chrisl | In PDF even a subset font is still "a complete font file" | 15:19.44 |
Robin_Watts | We'd just convert the subset and output another subset font. What's harder about that that dealing with a normal font? | 15:19.50 |
kens | Yes, a subset font is a font | 15:19.51 |
henrys | thane the output of any pcl ->. pdfwrite with text - TT subsets how does that go to SVG? | 15:20.15 |
kens | SVG with a TT font | 15:20.27 |
henrys | s/thane/take | 15:20.29 |
kens | The fact that its a subset is irrelevant | 15:20.36 |
| I'm assuming SVG has embedded font capability | 15:20.54 |
chrisl | It's just a font with fewer glyphs in it than the "source" font - still a complete font file | 15:21.07 |
Robin_Watts | AIUI, a subset font = a font where certain glyphs have been omitted. | 15:21.12 |
| Is an "incremental font" a PCL thing, where the definitions of each glyph are sent along as they are needed? | 15:21.32 |
kens | No an incremental font is where bits of the font (glyphs) are only downloaded as required | 15:21.50 |
| But its in PostScript mostly# | 15:22.00 |
Robin_Watts | If so, presumably pdfwrite does the hard work there by collating the definitions and outputting the fonts whole at the end. | 15:22.04 |
kens | For PostScript,/PCL yes | 15:22.17 |
henrys | I think pdf's definition of a whole font is quite different from what is needed by SVG | 15:22.30 |
Robin_Watts | kens: I don't see the distinction between your description of an incremental font and mine. | 15:22.34 |
kens | Robin_Watts : yes I misread yours | 15:22.41 |
Robin_Watts | ah, ok :) | 15:22.46 |
kens | henrys, a font is a font | 15:22.56 |
chrisl | henrys: if you pull a font stream out a PDF file, you can use it as a font file | 15:23.06 |
kens | As long as all the tables are correct, no TT interpreter can tell that its a subset | 15:23.16 |
chrisl | henrys: for example, is courier.ttf less of a "complete font" than msmincho.ttf because Courier has fewer glyphs? | 15:24.04 |
kens | Or an old COureier lacking a Euro glyph as compared to a new one which includes it ? | 15:24.31 |
chrisl | A better example, yes..... | 15:24.52 |
henrys | okay I was not aware you kept all the tables. | 15:24.52 |
Robin_Watts | henrys: For svg, we can have a font, with a font-face-uri that points to a font file. | 15:25.02 |
kens | pdfwrite (horrily) keeps some tables, you should regenarate them when creating a subset | 15:25.19 |
Robin_Watts | henrys: It's not clear to me quite what tables etc are available to us within mupdf; we currently have no way to get the original font data back from an fz_font. | 15:26.02 |
| That would need to be extended and to do so would probably require someone who understands fonts better than me to do it. | 15:26.25 |
kens | Robin_Watts : if you have the glyphs you can synthesise the rest | 15:26.45 |
Robin_Watts | kens: Currently we don't have the glyphs. | 15:27.02 |
kens | Oh, well you are going to need *something* from the font :-) | 15:27.15 |
chrisl | Given that PDF gives you an entire font stream, the best thing to do would be to expose that - in the TTF case, just write that out again | 15:27.30 |
kens | Unless you want to make a bitmap TT font (I assume you have the bitmaps at least....) | 15:27.34 |
Robin_Watts | Possibly I could extract the code from the draw device that can fallback to outlines and encapsulate that. | 15:27.36 |
| I am perfectly happy to have my high priority tasks as "jni bindings/svg output", as long as it's understood that I'll need to bother other people about the svg output bit of it :) | 15:28.27 |
henrys | I'm convinced now of resynthesis theory I'll wait and see for the practice ;-) | 15:28.42 |
paulgardiner | Access to the original font data from an fz_font would also allow more efficient generation of appearance streams via the pdf-write device. | 15:29.02 |
Robin_Watts | paulgardiner: What original data do you need? Just metrics, right? | 15:29.22 |
paulgardiner | For appearance streams, the most efficient would be a font dictionary reference because you tend to be creating the appearance stream for addition to a doc you have open | 15:30.58 |
Robin_Watts | paulgardiner: But you're not regenerating a font dictionary, right? | 15:31.33 |
| or are you thinking the general pdfwrite case rather than an annotation? | 15:32.04 |
| For an annotation, I'd imagine that you're working with a font that exists within the document. | 15:32.20 |
paulgardiner | Yes. That's the case I was refering to | 15:32.38 |
Robin_Watts | the annotation case? | 15:32.59 |
| what do you need from the font dictionary other than the metrics then ? | 15:33.28 |
tor7 | getting outlines out of the fz_font should be easy for non-type3 cases | 15:33.36 |
| it may be more robust to do that and write out a brand new TTF file than trying to use the actual font data | 15:34.03 |
Robin_Watts | It's possible that all we need is a "get me a font file for this fz_font" call. | 15:34.24 |
henrys | I guess it depends on Raed's usage but going to outlines or bitmaps gets everything including type 3's, you have a complete solution then you can go back and fill in supporting various font types as time permits. | 15:34.35 |
Robin_Watts | and that can either return the original or recreate one as appropriate. | 15:34.48 |
tor7 | Robin_Watts: that would be the easy end of the story. getting from glyph ids into something that an SVG interpreter will encode properly back to the same glyph id is harder. | 15:35.09 |
Robin_Watts | henrys: Right, but that's not entirely straightforward. I will look into it though. | 15:35.24 |
chrisl | tor7: if you degenerate to outlines, then reconstruct the font, you'll lose hinting...... | 15:35.49 |
tor7 | henrys: getting fonts from svg is a two step thing, I suspect. 1) as text + fonts, 2) as decomposed to primitives using another device or fallback mechanism | 15:36.01 |
paulgardiner | Robin_Watts: with annotations, referring to an existing font is the problem. I was able to get access to the metrics via existing means (for tt at least) | 15:36.58 |
henrys | tor7:I vote for rendering them as something and move on until we get requests. At least we have something working fast, but Robin_Watts says that is complicated also. | 15:37.24 |
Robin_Watts | paulgardiner: Right, so what do you need from the font struct other than the metrics. | 15:37.37 |
| s/./?/ | 15:37.48 |
paulgardiner | You've already found a font in the doc, or in some way created one (otherwise how did you create an fz_font object); you don't want the pdf device to resynthesise it | 15:38.07 |
tor7 | henrys: yes. a decomposing mechanism in mupdf is non-trivial. there are two ways I can see it being done. | 15:38.13 |
Robin_Watts | paulgardiner: Ah! You mean you need to know what the object number is for the font you are using. | 15:38.37 |
paulgardiner | Robin_Watts: yeah that might be sufficient (even ideal). I'm struggling to remember the details. | 15:39.14 |
tor7 | the one I'd like to see is inserting a "forwarding" device that takes one type of device call and decomposes into other calls, and just forwards the others. the tiling stuff may be a problem there with how the device communicates some information back to the interpreter. | 15:39.14 |
Robin_Watts | tor7: Hmm. Rather than having a 'forwarding' device... we currently have an fz_new_device call that sets up a device structure. All the devices use that then set the function calls | 15:40.09 |
| Possibly we should just have that return with the forwarding calls filled in already. | 15:40.27 |
paulgardiner | Robin_Watts: in some cases it may come down to just needing to know the local name of the font for use in the Tf operator | 15:40.32 |
tor7 | Robin_Watts: that might even be better :) | 15:40.41 |
Robin_Watts | paulgardiner: Right, but in some cases (like the fully general pdfwrite case), we need to know ALL the font details as we have to be able to output it. | 15:41.26 |
| And there are differences between an fz_font and an fz_font_desc - I think we might need the latter and only have the former in some cases. | 15:41.54 |
paulgardiner | Robin_Watts: Yes. I wasn't meaning to imply that was all we needed. It would just be nice to handle that case efficiently | 15:42.00 |
Robin_Watts | This is where my eyes start to glaze over. | 15:42.08 |
| and that's before the whole 'what code do we want to put where to mean a given unicode value' nightmare. | 15:42.50 |
kens | Robin_Watts : chrisl how can I tell Git to ignore the white space hook so I can put my changes to int.mak into a commit ? | 15:43.22 |
| anyone ? | 15:45.30 |
henrys | there was an interesting rumor floating around chicago that the Adode PDF print engine (PPE) no longer supported the internal PDF fonts everything had to be embedded⦠have to follow up on that. | 15:45.53 |
Robin_Watts | kens: You want tor7 for that. | 15:45.53 |
kens | tor7 how can I get Git to accept a commit with a tab in it ? | 15:46.08 |
chrisl | kens: git commit -n | 15:47.14 |
Robin_Watts | kens: is the problem when you are committing? Or when you are pushing? | 15:47.39 |
kens | Robin_Watts : committing | 15:47.48 |
Robin_Watts | If committing, then what chrisl said. | 15:47.48 |
kens | chrisl thanks, now I need to work out how to commit wihtout using Git GUI | 15:47.59 |
tor7 | kens: stage the changes with git-gui and then just "git commit -n" | 15:48.35 |
kens | aha, thanks | 15:48.42 |
Robin_Watts | tor7: Assuming he has an editor set up. Otherwise he'll be dumped into vi. | 15:49.00 |
kens | I'll cope | 15:49.09 |
chrisl | Being able to use vi should be mandatory - if you can't manage that, you're not worthy to use git..... ;-) | 15:50.00 |
Robin_Watts | ESC :wq! | 15:50.16 |
kens | VI sucks | 15:50.22 |
chrisl | It's no harder to use than Ghostscript! Oh, hang on....... | 15:50.53 |
Robin_Watts | chrisl: Oh yes it is. | 15:51.26 |
chrisl | I suspect people totally new to Ghostscript would disagree with you Robin_Watts | 15:52.07 |
Robin_Watts | A command line is more complex to use than a GUI, but is much more powerful. vi is more complex to use than a sane editor, but is... hmm... | 15:52.18 |
chrisl | Well, to be fair, I never claimed it was sane, nor good.... I don't actually like it, either | 15:52.53 |
| Hrm, sure there must be a way to give options to git commit via git gui...... | 15:54.19 |
kens | I couldn't find that one at least | 15:59.30 |
| well at least it built this time | 16:00.02 |
henrys | it would be much less work for me if you guys didn't finish your high priority projects ;-) | 16:00.31 |
chrisl | No, I can't see a way either | 16:00.31 |
Robin_Watts | henrys: Well, you just gave me svgout, right? :) | 16:01.34 |
kens | so, GS meeting ? | 16:02.29 |
mvrhel_laptop | i am back | 16:02.56 |
Robin_Watts | mvrhel_laptop: In your absence we gave you all the nasty jobs. | 16:03.21 |
mvrhel_laptop | ha | 16:03.35 |
henrys | the only thing I had for the meeting was to update everyone's project on the agenda | 16:04.24 |
| chrisl is the same | 16:04.40 |
chrisl | Yes, I'm swamped with other sh*t just now.... :-( | 16:05.03 |
henrys | mvrhel_laptop: it looks like your project is done? | 16:05.40 |
mvrhel_laptop | which project... | 16:06.00 |
henrys | we are reviewing the high priority projects in the agenda. | 16:06.32 |
Robin_Watts | mvrhel_laptop: "Color management in high level output devices." | 16:06.46 |
mvrhel_laptop | no. I still have much to do on that one. I did fix the GS color bugs | 16:07.04 |
Robin_Watts | and "MuPDF based windows viewer (phone priority) and GS color bugs discussed at meetings." | 16:07.12 |
mvrhel_laptop | discussed at meetings | 16:07.14 |
| still working on windows phone and also need to do windows desktop | 16:07.42 |
henrys | mvrhel_laptop: do you need the link emailed? | 16:08.12 |
mvrhel_laptop | the past week I have been trying to get rid of my p1 bugs and fix the bugs that were listed for me at the meeting | 16:08.12 |
| I have the workflow in front of me | 16:08.25 |
| so MuPDF based windows viewer (phone priority) and GS color bugs discussed at meetings. | 16:08.35 |
henrys | kens:what shall I put down for you? | 16:08.44 |
| kens:funny I don't think I ever had you on the list. | 16:09.16 |
mvrhel_laptop | henry: so windows 8 phone mupdf viewer, windows desktop gsview will be top of my list | 16:09.20 |
| and Color management in high level output devices is still there. I need to get V2 profiles out to kens | 16:09.37 |
kens | henrys I was hoping you wouldn't notice | 16:09.55 |
ray_laptop | Since we have confidential information on the workflowy page, shouldn't we have password login (at least), rather than relying on the URL being "secret" ? | 16:09.56 |
mvrhel_laptop | That should not take too long to do since I was just in that part of the code | 16:10.03 |
Robin_Watts | ray_laptop: The only way we could do that is if we all sign up to workflowy, then it can be a private workflowy that henrys 'shares' with each of us. | 16:10.33 |
mvrhel_laptop | that seems reasonable | 16:11.01 |
henrys | kens:well you're doing the color management business with michael yes? | 16:11.24 |
kens | henrys yes, still doing that (slowly) | 16:11.51 |
mvrhel_laptop | speaking of which, kens, you should be good to go with the V4 profiles that we now generate | 16:12.24 |
henrys | ray_laptop: I'm going to leave you project alone. | 16:12.33 |
mvrhel_laptop | I will try to get V2 working this week | 16:12.35 |
kens | mvrhel_laptop : I guessed that from the commits you had made, but I've3 beentoo busy to try it, I will make some time this week | 16:12.47 |
henrys | I think the list is okay anything else for the meeting? | 16:14.28 |
chrisl | henrys: how far back do your dealings with URW go? | 16:14.44 |
henrys | ray_laptop: I'm satisfied with the security it is https right? | 16:14.50 |
Robin_Watts | henrys: https means nothing if the URL is compromised. | 16:15.10 |
henrys | Robin_Watts: I know. | 16:16.30 |
marcosw | I'm about to enter the new fuzzing bugs I discussed at the meeting; so if everyone could go through the existing fuzzing bugs and grab the ones in their area of responsibility that would be great. | 16:17.04 |
henrys | marcosw: fair enough | 16:19.03 |
ray_laptop | marcosw: so, we should review the bugs with 'fuzz' in the subject that are assigned to support, right ? | 16:20.03 |
kens | how do we find the list ? | 16:20.04 |
henrys | I'm okay with the security as it is, if you guys feel strongly about it let me know and we'll work out something. | 16:20.24 |
ray_laptop | kens: for what I mentioned, I used: http://bugs.ghostscript.com/buglist.cgi?f1=OP&emailtype2=substring&list_id=9225&short_desc=fuzz&bug_severity=blocker&bug_severity=critical&bug_severity=major&bug_severity=normal&bug_severity=minor&bug_severity=trivial&bug_severity=enhancement&f0=OP&f4=CP&query_format=advanced&j1=OR&f3=CP&bug_status=UNCONFIRMED&bug_status=CONFIRMED&bug_status=IN_PROGRESS&email2=s | 16:20.50 |
| upport%40artifex.com&short_desc_type=allwordssubstr&emailassigned_to2=1 | 16:20.51 |
kens | :-) | 16:21.01 |
ray_laptop | oops. URL too long. | 16:21.01 |
henrys | kens:you look at the support bugs. | 16:21.10 |
ray_laptop | henrys: just support gets a lot of other bugs | 16:22.00 |
marcosw | ray_laptop: http://tinyurl.com/n9nfhhg | 16:22.13 |
ray_laptop | henrys: there are 240 bugs assigned to support. only 85 if I add "fuzz" in the Subject | 16:22.49 |
henrys | well support and not mupdf | 16:22.53 |
| right? | 16:23.00 |
ray_laptop | and only 73 with marcosw's URL :-) | 16:23.13 |
henrys | anything more for the meeting? | 16:23.42 |
Robin_Watts | Why are the fuzzing bugs bountiable? | 16:24.06 |
ray_laptop | marcosw: thanks. I'll go through those today and look for them | 16:24.18 |
henrys | let's call it, I'm off to do flood cleanup | 16:24.26 |
marcosw | henrys: ray_laptop brings up a good point, what to do with the mupdf bugs assigned to support (they were previously assigned to tor7 but at the meeting we decided that was sub-optimal). | 16:24.29 |
Robin_Watts | henrys: Good luck. | 16:24.42 |
| marcosw: Can we have a "mupdf-bugs@artifex.com" that they can be assigned to? | 16:25.20 |
chrisl | ray_laptop: I'm taking the fuzzing bugs that look like garbage collection and leaving the other memory manager ones for, I assume, you - is that okay, or should I take them, too? | 16:25.57 |
ray_laptop | or mupdf-support | 16:26.07 |
marcosw | Robin_Watts: I was thinking of assigning them all to you :-) | 16:26.25 |
ray_laptop | chrisl: that sounds OK to me. | 16:26.26 |
Robin_Watts | marcosw: I can ignore them there too :) | 16:26.37 |
chrisl | ray_laptop: cool. And if I miss some that are actually GC ones, you can punt them to me | 16:26.57 |
ray_laptop | chrisl: I'm just grateful to you for taking the GC ones. | 16:27.01 |
marcosw | They should be assigned to support@mupdf.com, but I don't think we have mupdf.com email addresses. | 16:27.33 |
ray_laptop | chrisl: gladly | 16:27.33 |
chrisl | ray_laptop: these are fuzzing bugs - I'm sure you're just as capable of ignoring them as I am ;-) | 16:27.34 |
henrys | anyway adjourned I'll be back in a couple hours. | 16:29.04 |
ray_laptop | on a positive note, I left the 2950 (poweredge) running memtest all night and it completed 6 passes. The only hits were the single glitch sequence I got early in the first pass. I'm going to run a test with ECC enabled next (since I'm still waiting for the drive caddies) | 16:29.44 |
Robin_Watts | ray_laptop: So subsequent passes didn't give the error? Or they gave the same errors? | 16:30.15 |
ray_laptop | Robin_Watts: no errors on any subsequent passes | 16:30.40 |
| chrisl: so, save / restore (isave.c) problems are mine, right ? | 16:34.37 |
chrisl | ray_laptop: Hmmm, good question. You can assign them to me (or I can grab them), if you prefer, I've dug around in that area before | 16:35.37 |
ray_laptop | chrisl: I'll skip them for now, and let you decide. I've also dug around in that area (as well as the GC), but fuzzing bugs aren't much fun. If you don't want them, just assign them to me. | 16:39.10 |
chrisl | ray_laptop: okay, I'll take them. Unfortunately, I think these, the GC ones and the memory manager ones aren't likely to get bounty attention :-( | 16:40.23 |
ray_laptop | chrisl: actually, since there are so many GC ones, why don't I take the (3 or 4) save restore ones. | 16:41.47 |
| chrisl: the GC ones take more time, I suspect. Just trying to balance the workload a bit | 16:42.22 |
chrisl | ray_laptop: Fine by me. I'm *really* hoping the GC problems will overlap, and they're not all separate issues! | 16:42.54 |
ray_laptop | chrisl: well, there is much to be said for blind optimism ;-) | 16:45.32 |
chrisl | ray_laptop: Why do think I was willing to look at GC problems in the first place! Too stupid to know better ;-) | 16:46.30 |
| I think I'm going bug-blind....... | 16:49.20 |
ray_laptop | I assume that kens will take the interpreter ones. Is that OK with you, kens (and of course the pdfwrite ones) | 16:49.33 |
kens | have done pdfwrite ones | 16:49.48 |
ray_laptop | OK, list is down to 45 remaining. I've taken the clist ones, and the non-GC memory ones (including save restore ones) | 16:51.31 |
| I only ended up with 13. :-) If I chip away at one a week, I'll probably be able to get them done by next staff meeting. Now to just remind myself to do that :-/ | 16:53.39 |
Robin_Watts | I've taken ones that look like mine. If anyone sees any others that seem like they should be mine, please say. | 16:54.18 |
kens | Well I *think* that's all mine, back up to 95 bugs :-( | 17:03.25 |
ray_laptop | Robin_Watts: did you take the image_render ones ? | 17:04.14 |
Robin_Watts | ray_laptop: I took at least one. The others might have been in the halftoning code, so I left them in case mvrhel wanted first dibs. | 17:04.52 |
mvrhel_laptop | wait where is this list that we are supposed to be looking at | 17:05.24 |
ray_laptop | Robin_Watts: yeah, I figured the image_render_ threshold stuff would be mvrhel_laptop's | 17:05.35 |
| mvrhel_laptop: http://tinyurl.com/n9nfhhg | 17:05.51 |
Robin_Watts | http://tinyurl.com/n9nfhhg | 17:05.59 |
kens | I'm heading off for the night, goodnight all | 17:06.01 |
Robin_Watts | night kens | 17:06.05 |
ray_laptop | nite, kens | 17:06.10 |
| marcosw: at least one of my bugs has a LOT of files. How can we re-test the fuzzing files after a fix (re: http://bugs.ghostscript.com/show_bug.cgi?id=694133&list_id=9248) | 17:12.31 |
mvrhel_laptop | ok so I think I have taken all of mine | 17:13.32 |
ray_laptop | henrys: you've marked some of my bugs as "Bountible to Shelly and Simon (only)" Does that mean I shouldn't work on them ? | 17:13.39 |
Robin_Watts | ray_laptop: Lots of them are marked thusly. | 17:13.51 |
mvrhel_laptop | so where again do I get the offending files? | 17:14.06 |
| Robin_Watts: ^^ | 17:14.09 |
ray_laptop | I'd be happy if they just take the openjpeg ones. | 17:14.22 |
marcosw | ray_laptop: after you have a fix for one or more of them you can assign them to me and I'll re-run them. If there are still issues I'll update the bug and assign it back to you. | 17:14.34 |
ray_laptop | mvrhel_laptop: tests_private/fuzzing | 17:14.35 |
Robin_Watts | mvrhel_laptop: what ray said. | 17:14.44 |
mvrhel_laptop | gotcha. | 17:14.55 |
marcosw | btw, everyone shouldn't worry about henrys being left out of the fuzzing party. I'm now fuzzing GhostPCL :-) | 17:15.24 |
mvrhel_laptop | trying to fix 693365 first then I will take a look at a few fuzzing bugs | 17:15.30 |
| yeah marcosw! | 17:15.43 |
ray_laptop | mvrhel_laptop: I get mine from peeves.ghostscript.com:/home/regression/tests_private/fuzzing | 17:16.49 |
mvrhel_laptop | 693365 is a little weird. the image code is inserting a clipper device between my push and pop of a transparency group for pattern rendering and that is wreaking havoc | 17:16.49 |
| ray_laptop: oh thanks | 17:17.02 |
marcosw | mvrhel_laptop: don't get too excited, perhaps all of the GhostPCL fuzzing issues will be in the color management code... | 17:17.06 |
mvrhel_laptop | that will be much better than fooling with svn | 17:17.10 |
| funny | 17:17.17 |
ray_laptop | mvrhel_laptop: agreed. | 17:17.21 |
mvrhel_laptop | but you are probably right marcosw | 17:17.44 |
ray_laptop | bbiab. Need more coffee... | 17:23.35 |
marcosw | I'm off to uni, back online later. | 17:29.28 |
Robin_Watts | Everyone should reload the ghostscript dashboard, please. | 17:46.38 |
| actually. don't. | 17:50.53 |
| It's not working for me, and it was earlier, and I can't see why. | 17:51.04 |
| oh, working again now. How strange. | 17:51.20 |
mvrhel_laptop | ugh. 693365 is a pain. clist only, resolution dependent, pattern transparency | 18:14.11 |
Robin_Watts | mvrhel_laptop: Can you test something for me please? | 18:27.06 |
| If you reload the dashboard, there should be a link on the right "Test files" | 18:27.18 |
mvrhel_laptop | sure hold on | 18:27.26 |
Robin_Watts | Follow that link, then enter the name of a test file from our repo, and it should download it to your computer. | 18:27.51 |
mvrhel_laptop | nice! | 18:27.56 |
Robin_Watts | I may try to hook it up into the reports etc at a later date. | 18:28.16 |
| bmpcmp is HTML so that's easy. | 18:28.32 |
| Other reports are just text currently, so I'd need to HTMLise them on the fly and put the links in. | 18:28.57 |
mvrhel_laptop | do the fuzzy files work? | 18:28.59 |
Robin_Watts | mvrhel_laptop: Should do. | 18:29.08 |
| Any file within tests or tests_private should work. | 18:29.16 |
mvrhel_laptop | so if I have 1677.pdf.asan.39.754.ps.pkmraw.300.0 | 18:29.18 |
| listed as a problem. | 18:29.44 |
Robin_Watts | You need the full path. | 18:29.45 |
| So tests_private/fuzzing/1677.pdf.asan.39.754 | 18:30.24 |
mvrhel_laptop | ok that worked | 18:30.43 |
Robin_Watts | cool, thanks. | 18:30.49 |
mvrhel_laptop | i.e. tests_private/fuzzing/So tests_private/fuzzing/1677.pdf.asan.39.754.ps | 18:31.05 |
Robin_Watts | ? | 18:31.34 |
mvrhel_laptop | you left off the ps | 18:31.44 |
Robin_Watts | I think it probably returns you zero length files if you give a duff name. | 18:31.55 |
| I should fix that. | 18:32.00 |
mvrhel_laptop | yes it does | 18:32.37 |
| back to fun with patterns/transparency/clist | 18:33.26 |
Robin_Watts | mvrhel_laptop: There is no .ps for that file, AIUI. | 18:34.38 |
mvrhel_laptop | ? | 18:34.52 |
| oh | 18:35.17 |
Robin_Watts | The file is called tests_private/fuzzing/1677.pdf.asan.39.754 | 18:35.38 |
| it was originally a file called 1677.pdf and it was fed through address sanitiser with settings 39 and 754. I think. | 18:36.06 |
mvrhel_laptop | oh ok | 18:36.14 |
Robin_Watts | I'm not sure where the .ps comes from - maybe that's marcosw saying it was a gs thing? | 18:36.24 |
mvrhel_laptop | that is a good question | 18:36.33 |
| and a bit confusing to me | 18:36.40 |
Robin_Watts | marcosw: ping? | 18:36.55 |
| mvrhel_laptop: OK, it now gives an error on a bad file rather than downloading a 0 byte file. | 18:39.32 |
mvrhel_laptop | great. very nice Robin_Watts | 18:45.39 |
| now I would like to understand marcosw file naming convention. but first this bug. it is looking like there are 2 issues | 18:46.16 |
Robin_Watts | The files, as they arrived from customer that did the fuzzing are: 1167.pdf.39.754 | 18:46.56 |
Gigs- | I need the attachments to http://bugs.ghostscript.com/show_bug.cgi?id=694600 marked private ASAP please | 18:47.03 |
Robin_Watts | marcosw then appends: ps.pkmraw.300.0 to it. | 18:47.19 |
| presumably cos it was using gs, and the pkmraw device at 300dpi with no banding. | 18:47.35 |
| Gigs-: Done, I think? | 18:48.17 |
mvrhel_laptop | I don't like ps used to denote gs really | 18:48.46 |
| but that is me | 18:49.03 |
Robin_Watts | mvrhel_laptop: I'm *guessing* that it's ps/pcl/xps etc? | 18:49.30 |
| but I'd like to hear from marcosw to be sure. | 18:49.39 |
Gigs- | Robin_Watts: yeah it looks like it, thanks | 18:50.04 |
Robin_Watts | Gigs-: No worries. | 18:50.16 |
Gigs- | is the current gs git somewhat stable | 18:50.32 |
Robin_Watts | Gigs-: I believe so, why? | 18:51.04 |
Gigs- | that bug was on a january git snapshot, it may not reproduce on a newer | 18:51.20 |
Robin_Watts | No major new features have been dropped in to destabilise it recently. | 18:51.28 |
Gigs- | if you have a minute to run that command line on it and see if you can reproduce | 18:51.36 |
Robin_Watts | Gigs-: 52Meg download is a problem for me at the moment, sorry. | 18:52.29 |
Gigs- | ok no problem | 18:52.51 |
Robin_Watts | ADSL is broken - connected through 3G dongle with a rigid fair use policy :( | 18:52.57 |
Gigs- | at least it's 3g | 18:53.22 |
Robin_Watts | Gigs: Indeed. Other than the fair use thing, it's quite usable. | 18:54.10 |
| Certainly it's better than struggling along with no connection at all. | 18:54.23 |
Gigs- | I had to fight with satellite for a long time at home until I got lucky with cable | 18:54.38 |
mvrhel_laptop | looks like tile_rect_trans_simple is overrunning a buffer :( | 18:54.39 |
Gigs- | there's no cable service anywhere around me but I'm somewhat close to an elementary school that had fiber run through a grant, so there's this little island of cable service | 18:55.03 |
Robin_Watts | mvrhel_laptop: Oh, well that code is a piece of cake. Transparency, clist, patterns, all simple stuff :) | 18:55.10 |
| If you had garbage collection in there, it'd be a full house :) | 18:55.31 |
Gigs- | are you talking about mine or the pkm? | 18:55.34 |
mvrhel_laptop | exactly | 18:55.40 |
Robin_Watts | mvrhel's problem. | 18:55.43 |
Gigs- | k | 18:55.45 |
Robin_Watts | Gigs: I live about 400 yards from the exchange, and about the same from the local school. Still no cable here :( | 18:56.36 |
Gigs- | well at least you can probably get good DSL speeds when that's working at that distance | 18:56.54 |
Robin_Watts | The exchange maxes out at 8mbps down and 448kbps up. | 18:57.20 |
| and that's what I get (when it's not broken) | 18:57.30 |
Gigs- | not too bad | 18:57.33 |
Robin_Watts | certainly way better than I used to get, yes. | 18:57.42 |
Gigs- | that's so asymmetrical you might run into ack starvation limiting your down though | 18:58.09 |
| though TCP has gotten much much better about agressively opening up the window | 18:58.40 |
| actually I can't safely upgrade gs until http://bugs.ghostscript.com/show_bug.cgi?id=693447 is fixed | 19:03.40 |
| that breaks one of my internal regression testing script | 19:03.59 |
| I just took the risk when I upgraded in January | 19:04.28 |
mvrhel_laptop | checking blame on this one and it looks like I am stuff with the problem | 19:06.41 |
| s/stuff/stuck/ | 19:06.47 |
| time for some specialize debug printing to be done for this issue | 19:23.08 |
ray_laptop | all: Shailesh is "Shelly", right ? | 20:15.54 |
| so what's the common IRC nicks here ? | 20:16.44 |
| I mention it because I just took assignment of bug 694132 this morning, and a commit with a partial fix was just made. I don't want to duplicate effort and want to make sure that our "bounty" fixers know to assign bugs to themselves BEFORE working on them | 20:18.36 |
| hopefully, they can also change status to "IN PROGRESS" as well when they actually start working on them (as we do) | 20:19.53 |
| or are supposed to do. | 20:20.05 |
| Thesse fuzzing bugs are a pain. Of the three that I've looked at, they all "work for me". Have to try next on 64-bit peeves... | 20:25.10 |
mvrhel_laptop | caught this thing with it hand in the cookie jar | 20:43.55 |
henrys | back from armageddon cleanup - some folks had dead fish in their basement from the flood - not aquarium fish | 20:56.17 |
| for the logs leaving them bountiable is fine, ray_laptop, but if they are in difficult code I think we should work on them as time permits | 20:58.14 |
b0fh_ua | Hello! I cropped PDF file and now it's page size is a bit weird, how can I reset page size of cropped file back to A4? | 21:34.43 |
| Forward 1 day (to 2013/09/18)>>> | |