| <<<Back 1 day (to 2015/02/02) | 20150203 |
rayjj | Finally have some sensible (mostly) numbers for gs and pcl6 RAM usage. There are a few PDf files that seem "out of the park" on RAM that need investigation, but out of 10,000+ ATS pages that's not bad | 07:39.34 |
| I sent (to tech) an email with the .csv files that hopefully allows you to sort and/or grep to get the data you want. Excel loads it, so SO should as well :-) | 07:41.09 |
| time to sleep before the meeting in ~ 8 hrs. | 07:42.11 |
clmates | hi | 10:47.05 |
ghostbot | Welcome to #ghostscript, the channel for Ghostscript and MuPDF. If you have a question, please ask it, don't ask to ask it. Do be prepared to wait for a reply as devs will check the logs and reply when they come on line. | 10:47.05 |
clmates | I have a question regarding the usage of ps2pdf.bat with concurrent calls | 10:48.04 |
| I'm using ghostscript in order to generate pdf output from a reporting system | 10:48.30 |
| this system can hadle several reports at the same time so it can call ps2pdf.bat concurrently for several reports | 10:49.03 |
| I'm having problems to generate the pdf sometimes, and I think this is because the _at and _at2 files being generated in TEMP folder for all calls, so they get locked or mixed | 10:50.03 |
chrisl | clmates: it won't work - don't use ps2pdf.bat just call Ghostscript directly and that should work just fine | 10:50.23 |
kens | chrisl beat me to it | 10:50.37 |
clmates | so my question is if this is possible to make these bat files to generate inside TEMP a subfolder with the process number $$ | 10:50.47 |
| so each call has his own temp | 10:51.04 |
kens | You can alter the batch files if you insist on doing it that way, but as chrisl says, don;t bother | 10:51.15 |
clmates | and then remove this process folder at exit | 10:51.18 |
chrisl | I refer my esteemed colleague to the answer I gave earlier..... ;-) | 10:51.42 |
clmates | just another question then, because I'm not sure how to integrate with the rport software with direct call , sorry | 10:52.17 |
kens | If you can call a .bat file tehn you can call a .exe | 10:52.46 |
chrisl | clmates: so you'll call gswin32/64c.exe (instead of ps2pdf.bat) and use the parameters: -sDEVICE=pdfwrite -dCompatibilityLevel=1.4 -o <output file>.pdf <input file> | 10:54.07 |
kens | I'd dump the COMpatibilityLevel, if you use 1.4 it limits 'something', I forget what, but I recently changed the default level to 1.5 | 10:54.44 |
| I didn't bother changing the scripts | 10:54.59 |
clmates | ok, I will try that, many thanks | 10:55.30 |
chrisl | NP | 10:55.36 |
tor8 | Robin_Watts: two commits on tor/master | 11:12.55 |
| the mujs bugfix and the new URW fonts | 11:13.02 |
Robin_Watts | So we have no CJK fallback? | 11:14.32 |
tor8 | no change for the CJK fallback | 11:14.45 |
| this is only affecting the non-CJK fallbacks | 11:14.51 |
Robin_Watts | Gotcha. | 11:14.58 |
| lgtm. | 11:15.01 |
tor8 | thousands of regression diffs | 11:15.09 |
Robin_Watts | of course. | 11:15.17 |
tor8 | most due to the replacement of droid sans with the base14 as a substitute | 11:15.25 |
| but also several glyphs that have changed shapes slightly in the new URW fonts | 11:15.38 |
| Robin_Watts: thanks. pushed. | 11:15.45 |
| I've been messing with the embedded context in fz_document and the *_document subclasses | 11:17.19 |
| not really come to a compelling solution | 11:17.29 |
| we obviously want to keep the pdf_document, xps_document etc as fz_context substitutes | 11:17.57 |
| or at least that's what I've assumed | 11:18.43 |
| currently I've got the fz_ functions calling the rebind function and then calling the real callback | 11:20.15 |
Robin_Watts | Hmm. So we'd still bind context to doc? | 11:20.16 |
| right. | 11:20.28 |
tor8 | another approach would be to do the rebind inside a pdf_ wrapper function | 11:20.49 |
| and a third approach would be to not embed the fz_context in the document types at all | 11:21.02 |
| as my drop branch stands at the moment, the document structs are the only remaining fz_context embedders | 11:21.30 |
Robin_Watts | I might be tempted to use explicit contexts for fz_docs too. | 11:22.05 |
| Consistency etc. | 11:22.17 |
tor8 | the pdf_op_processor stuff is a bit messy with all the structs around it, but that's not as externally visible so I'm saving that for last | 11:22.19 |
| yeah. if we're breaking a lot of APIs with the next release, might as well go the full way | 11:22.34 |
| I'll take a stab at that after lunch. thanks. it might turn out better than I expected. | 11:23.10 |
| if we keep the pdf_doc as a context-type, the split with pdf_ functions taking a ctx and a doc will drive me nuts | 11:23.36 |
vtorri_ | hey | 11:42.47 |
ghostbot | Welcome to #ghostscript, the channel for Ghostscript and MuPDF. If you have a question, please ask it, don't ask to ask it. Do be prepared to wait for a reply as devs will check the logs and reply when they come on line. | 11:42.47 |
vtorri_ | i'm aplnning to udate mupdf from 1.2 to 1.6 | 11:42.59 |
| there are a lot of changes | 11:43.04 |
| is there some doc about how to update ? | 11:43.14 |
| i remember that in some releases, there was a script which changed the names from a version to another | 11:43.44 |
| but i can't find that script anymore in mupdf 1.6 | 11:43.58 |
clmates | chrisl just to confirm that this seems to work much better, I'm not getting any kind of error about locked files and also seems to be faster, many thanks again | 11:52.10 |
tor8 | vtorri_: might be better if you wait for 1.7 | 11:52.31 |
chrisl | clmates: cool, glad to help | 11:52.36 |
tor8 | there are significant changes incoming | 11:52.36 |
Robin_Watts | tor8: Are you intending to get your drop stuff into 1.7? | 11:54.08 |
vtorri_ | tor8: ok, thank you | 11:54.08 |
| tor8: any idea when 1.7 will be out ? | 11:54.23 |
Robin_Watts | This month, or early next month, I think. | 11:54.51 |
vtorri_ | ok, thank you | 11:55.03 |
tor8 | Robin_Watts: I am still hopeful. | 12:21.12 |
| Robin_Watts: most of the changes are pretty mechanical | 12:21.31 |
Robin_Watts | coo. Found an interesting tool, and look at the author... | 13:21.23 |
| https://code.google.com/p/noto/source/browse/third_party/fontcrunch/README.md | 13:21.25 |
tor8 | Robin_Watts: ooh. that'd be interesting to try out on our new URW fonts. | 13:22.34 |
kens | Err, TrueType outlines, our new fotns are CFF outlines | 13:22.51 |
Robin_Watts | That works on ttf, which are ... what kens said. | 13:22.58 |
tor8 | kens: we've got them in TTF as well, no? | 13:23.04 |
kens | has no idea | 13:23.10 |
| We have old ones in TTF | 13:23.15 |
Robin_Watts | tor8: We really want to only be standardising on 1 format. | 13:23.28 |
kens | I htought Henry wanted 'a singel deliverable' so I assumed we just got OTF with CFF outlines | 13:23.37 |
Robin_Watts | Possibly extending the code to work with cubics rather than quads would be possible. | 13:23.49 |
tor8 | Robin_Watts: kens: I just assumed based on the Artifex_update.zip in henrys home directory that contains TTFs of the new base-12 fonts with greek and cyrillic glyphs | 13:25.05 |
| but I agree, OTF for one deliverable is ideal | 13:25.17 |
kens | didn't know that | 13:25.18 |
tor8 | kens: the OTF is half the size of the TTF though | 13:26.22 |
| okay, so try it on the droid sans fallback font? | 13:26.39 |
kens | Yeah, I think we established that the CFF outlines were the smallest possible set. | 13:26.40 |
| It would be interesting to try it on DroidSans | 13:26.51 |
Robin_Watts | tor8: That sounds a good idea. | 13:26.54 |
tor8 | kens: I've extracted the CFF table from the OTF and used that for MuPDF | 13:26.55 |
Robin_Watts | but then you might assume that they've already done that :) | 13:27.04 |
tor8 | saves us a few 100k of pointless 'cmap' table cruft that mupdf doesn't use anyway | 13:27.10 |
kens | tor8 yeah, but we'll need to keep them as OTF for GS, in order to share them with GhostPCL, unless CHris decides that we can add CFF loading to GhostPCL instead. | 13:27.54 |
Robin_Watts | tor8: Is that CMAP cruft used by gs and/or pcl ? | 13:30.26 |
kens | I think Zoltan can forget his latest request. No way I'm changing setpagedevice to work differntly to the specification | 13:30.47 |
tor8 | Robin_Watts: I'd assume it's used by PCL | 13:30.48 |
| mupdf uses the glyph names to build its encoding tables | 13:31.04 |
Robin_Watts | If not, we could strip it out of the otfs. | 13:31.08 |
| fair enough. | 13:31.14 |
tor8 | if you strip out the cmap from the OTF you've pretty much just got a raw CFF :) | 13:31.30 |
| the 'hmtx' table is pure redundancy | 13:32.21 |
| the 'cmap' in the OTF may actually add something of value, as well as the kerning and GPOS etc tables | 13:32.40 |
kens | CMAP is used by GS I believe, when using a TT as a font substitute | 13:32.48 |
| Of course, that's not hte same as OTF+CFF | 13:33.28 |
chrisl | kens: the thing from Zoltan.... couldn't he get what he needs by calling putdeviceparams directly? | 14:38.45 |
kens | No, I don't think so, he wants it to track gsave/grestore | 14:39.18 |
Robin_Watts | If he needs to track gsave/grestore, then he needs to understand gc. | 14:39.39 |
kens | Since the device parameters are not affected by grestore, it won't work | 14:39.44 |
Robin_Watts | and if he really understood gc, I suspect he'd just amend the graphics state itself. | 14:39.59 |
kens | Robin_Watts : I believe he needs to store his params in his own dictioanry, and hook grestore so that if any of the parameters have changed he updates his device. | 14:40.17 |
chrisl | With a bit of Postscript hacking, it could be made to work | 14:40.38 |
Robin_Watts | Possibly we should mail him back and get him to tell us exactly what he wants to achieve. | 14:41.04 |
kens | I understand what he wants, I believe, I don't think his method is a good plan. | 14:41.25 |
| Breaking setpagdevice to achieve his goals is a Bad Idea, especially the way our setpagedevice is implemented | 14:41.43 |
| But its the wrong thing to do anyway | 14:41.55 |
| As you said, he could do it by altering the graphics state | 14:42.20 |
| But that would be hard, and would need to be kept up to date, he doesn't want to have to do that | 14:42.37 |
chrisl | But using putdeviceparams should get him the communication with the device without erasing the page, and a bit Postscript hacking would get the parameters obeying gsave/grestore etc | 14:42.47 |
kens | chrisl that's what I'm suggesting above | 14:43.04 |
| He needs to store his parameteres in a dict, giving him save/restore, and hook grestore so that if his params have changed, he sends them to the device | 14:43.29 |
| Ideally he would want to store his redefinition in systemdict of course | 14:44.20 |
chrisl | Yeh, I just felt that if we point him at putdeviceparams that gives him the (main) tool he needs to get where he wants | 14:44.23 |
kens | I think he already knows that exists, from his mail | 14:44.35 |
| When I get some time I was going to have a go at implementing it, but I'm lost in a twisty maze at the moment | 14:45.58 |
chrisl | TBH, if he knows about putdeviceparams he has everything he needs to do what he wants, and in a way that won't need to updated regularly with new gs releases | 14:47.11 |
kens | Yeah, I just wanted to present him with a solution, when I tell him to get knotted | 14:47.33 |
Robin_Watts | The new noto fonts from google weigh in at 15Meg+ for CJKV. | 14:49.17 |
| Anyone want to guess what the bitstream fonts weighed in at for the same? | 14:49.33 |
chrisl | Does bitstream use it's own format? | 14:51.50 |
Robin_Watts | yes. | 14:52.36 |
| uses strokes to make glyphs. | 14:52.49 |
chrisl | Could be any size then....... | 14:53.06 |
Robin_Watts | Not great for print quality, but quite good for screen use. | 14:53.09 |
| less than 1Meg. | 14:53.11 |
chrisl | I bet the shapes are pretty poor - like the Microtype fonts..... | 14:54.06 |
kens | Well, I think I just came across a major problem with pdfwrite and the device subclassing. pdfwrite maintains a pointer to the device in its font_cache elements. If we move the location of pdfwrite around, then the pointer to the device will be invalidated. It works OK for the first/last page code, because that gets installed right up front before pdfwrite sees any fonts, and never gets removed. | 14:54.34 |
Robin_Watts | kens: Can you have them keep a pointer to something that doesn't move, and that contains a pointer that does? | 14:58.27 |
kens2 | Robin_Watts : I think I can just do without the pointer anyway, it seems to be slightly pointless. I can pass in the device pointer from the callingfunction. | 15:01.41 |
Robin_Watts | even better. | 15:01.51 |
kens2 | WHich will also save (very slightly) on storage too | 15:01.54 |
| I'm just trying to figure out if that will work right now | 15:02.13 |
| Being pdfwrite its ahrd to tell | 15:02.24 |
| Yeah there's even a comment tghat says we only use it for removing the font cache element, Dumb! | 15:03.15 |
tor8 | Robin_Watts: are these the Noto Han fonts that adobe made? | 15:06.25 |
Robin_Watts | Yes. | 15:06.40 |
tor8 | they're pretty nice, and have variants for all glyphs that differ between traditional, simplified, korean and japanese scripts | 15:07.07 |
Robin_Watts | yeah. I'm having problems with them in SOT at the moment, which I think is due to stuff not working properly for fallback. | 15:07.40 |
tor8 | I have considered using them, since we know which script to select based on the CID ordering used | 15:07.45 |
| but they're pretty massive compared to DroidSansFallback | 15:08.05 |
Robin_Watts | yes. | 15:08.20 |
tor8 | we might consider creating a subset of glyphs based on the most common code pages and see how small we can get that | 15:08.57 |
kens2 | OK turns out that there are a slew of other places that pdfwrite holds on to a copy of the device pointer, and in a lot of them, it needs too. Basically this isn't going to work I don't think. | 15:20.15 |
| It'll be OK for devices that get pushed at the start and persist, but that's all. | 15:20.35 |
henrys | nothing like a little zoltan ;-) | 15:21.57 |
| meeting in 10 minutes | 15:22.06 |
rayjj | kens2: you could just make pdfwrite "immovable" -- it's probably just about large enough to usually be in its own chunk anyway | 15:32.18 |
marcosw | morning | 15:32.38 |
kens2 | rayjj I can't do that, the whoel method of working for the device subclassing involves copying devcies around | 15:32.50 |
fredross-perry | morning | 15:32.53 |
henrys | if everyone could take a look at rayjj email and weigh in if needed say in the next 2 hours that would be great. We need to get that out. | 15:33.01 |
rayjj | henrys: the concern is what to say about the files that take a lot of RAM -- and not just clist. | 15:34.41 |
henrys | kens2: have you had a chance to look at the performance issue with forms, enough so we could give scheduling information or tell them what parts we intend to address, if any? | 15:35.52 |
kens2 | henrys, I have done nothign further with it, all my thoughts are in the bug report. Its up to you if you want me to look into it. We can do one of 3 things: | 15:36.39 |
| 1) DO nothing | 15:36.42 |
| 2) Implement a full form cache | 15:36.50 |
| 3) Implement a form cache look-alike for high level devices | 15:37.05 |
| 1 is easy | 15:37.11 |
mvrhel_laptop | morning. sorry I am late | 15:37.21 |
kens2 | 2 would take a long time, and probably has little benefit | 15:37.24 |
tor8 | henrys: which exact license do we have for the new URW++ fonts? | 15:37.27 |
henrys | kens2: okay I got the impression there were minor improvements that could be done. | 15:37.36 |
kens2 | 3 I;m guessing would be a couple of days work, and would only work for high level devices | 15:37.41 |
tor8 | and is it the same license for the OTF and the TTF formats? | 15:37.49 |
chrisl | rayjj: do those aberrant PDFs share any common characteristics? (transparency, shadings etc)? | 15:38.15 |
henrys | GPL or commercial only for the 3 fonts | 15:38.24 |
rayjj | kens2: for 3, does the resulting high level output render faster ? | 15:38.36 |
tor8 | which version of GPL? 2 or 3? | 15:38.37 |
kens2 | henrys I cna fix the problems with the /Pattern colour space and pdfwrite. That would resolve the multiple form caching and ouwl get a smaller output file, but would not address the performance | 15:38.55 |
| rayjj yes, that one gets performance benefits, for high level devices only | 15:39.15 |
chrisl | henrys: how do those OTFs differ from the Type 1s from a couple of weeks ago? (other than being OTF) | 15:39.41 |
rayjj | kens2: I meant, does rendering the resulting PDF (created from the #3 changes) get any benefit | 15:40.22 |
henrys | tor8: good question ... they know nothing about the GPL, they want it to be the same as the fonts they gave last to the community and this is really an upgrade for those fonts - so that must be the original GPL given the timeframe. | 15:40.23 |
kens2 | rayjj no, I wouldn't think so, we would already have spotted the duplicatws and removed them. | 15:40.43 |
| Note that this is something I haven't yet tested, due to the problems with the /Pattern colour space definition | 15:41.04 |
tor8 | henrys: I'll drop in the GPL2 license file then. I'm planning to upload these font files to a separate repository for the community to use. | 15:41.10 |
rayjj | kens2: that's what I thought, I just wanted to confirm it | 15:41.11 |
henrys | chrisl: there intended to be compatible but I believe you shouldn't have to fool with encoding to reach all the glyphs, right? Weren't the type 1's just a standard | 15:41.23 |
| ? | 15:41.25 |
kens2 | You can only have 255 glyphs in an Encoding. Type 1 fonts may contain more glyphs and you have to change the ENcoding to get to them | 15:41.54 |
| StandardEncoding is just a default | 15:42.10 |
chrisl | That'll be true of OTF/CFF as far as Ghostcript is concerned | 15:42.25 |
henrys | kens2: right the type 1 fonts should have exceed the 255 | 15:42.26 |
rayjj | henrys: the whole point of having Encodings is for other languages to pull in / exclude glyphs needed by a language (into the base 255 or low 127) | 15:43.11 |
henrys | chrisl: the beziers and metrics should be identical. | 15:43.16 |
chrisl | henrys: so we expect the last lot of Type 1 fonts we got to have the new glyphs in them, too? | 15:43.27 |
henrys | chrisl: yes | 15:43.36 |
chrisl | Cool, I don't need to do anything more for now, then | 15:43.50 |
rayjj | chrisl: to your earlier question (that I missed): I haven't looked at the aberrant files yet. I've been focused on collecting the data for now | 15:45.08 |
henrys | chrisl: I'd like to get them in the build soon is there some reason not to? | 15:45.41 |
chrisl | henrys: which? | 15:45.53 |
henrys | chrisl: the 3 fonts we just received | 15:46.09 |
chrisl | henrys: the Type 1 versions can go in immediately | 15:46.24 |
rayjj | I found it interesting/alarming that tests_private/pdf/PDF_1.7_ATS/AIX361DC_Save.pdf only used 40Mb of compressed clist, but needed 850Mb of RAM | 15:46.45 |
kens2 | henrys ps2write won't pass all the CET tests with these fonts in CFF format | 15:46.51 |
henrys | chrisl: no I meant the others. | 15:47.05 |
chrisl | There's no way I'm making that kind of change a month before a release | 15:47.21 |
henrys | 3 fonts geez .... okay | 15:47.58 |
| with what kens2 is saying though right... I didn't realize that. | 15:48.24 |
chrisl | Er, the three most often used of the built in fonts..... | 15:48.33 |
henrys | chrisl: tor8 just checked in his ;-) | 15:49.01 |
kens2 | I put in a quick hack to avoid the seg fault, I didn't attempt to fix the problem properly, its several more days work to address it | 15:49.02 |
tor8 | in chrisl's defence, they are 3 pretty crucial fonts... | 15:49.59 |
henrys | rayjj: then we need different files for this report. | 15:50.17 |
| rayjj: and bugs for the memory issues | 15:50.38 |
chrisl | tor8: and Ghostscript and Postscript can do considerably more heinous things with fonts that mupdf would ever do...... | 15:51.04 |
tor8 | chrisl: exactly | 15:51.19 |
| I'm pretty confident nothing will break badly with mupdf with these fonts | 15:51.35 |
chrisl | henrys: we also have a small issue to decide what to do about: we are the de-facto upstream source for the URW fonts: we *will* get complaints if we stop "shipping" the Type 1 versions | 15:52.27 |
henrys | chrisl: well a clusterpush and a bug with results would be good. | 15:52.49 |
chrisl | henrys: the only problem is the ps2write one - I've fixed the others I found (on a branch) | 15:53.19 |
henrys | we do have a potential customer, the customer has licensed these fonts we need to be ready. | 15:53.58 |
rayjj | henrys: What files do you suggest ? The JEITA files are reasonable (the worst case is J12 which is < 40Mb | 15:54.11 |
kens2 | suspects the customer doesn't need ps2write | 15:54.21 |
henrys | kens2: yes that is a very good point. | 15:54.33 |
chrisl | henrys: throwing a release over the wall without proper testing does not count as "being prepared" | 15:54.47 |
rayjj | chrisl: are you sure we will get complaints? The free users almost all use freetype, so why would they demand Type 1 vs OTF ? | 15:55.40 |
kens2 | Of course, if we are to ship OTF-CFF fonts as standard, then the weird case in ps2write (TYpe 0 font with two descendants, one type 3 and one built-in font) *does* need to be fixed before we do a release with these fonts built-in | 15:55.47 |
| rayjj someone will complain..... | 15:56.08 |
| Pretty much guaranteed | 15:56.17 |
rayjj | kens2: yeah, true. Some people just live to complain | 15:56.39 |
henrys | I guess I didn't perceive the change as that great a deal if it passes the cluster test. and yes somebody will complain ... | 15:56.43 |
chrisl | rayjj: I'm sure - I've had complaints that glyph names have changed in the charstrings dictionary, that stuff in the FontInfo dictionary has changed etc | 15:56.59 |
kens2 | henrys I'd be reluctant to do such a major change just before a release. A cluster test is great, but its far from exhaustive. | 15:57.20 |
henrys | sorry to take up the whole meeting with this again but it is rather important. | 15:57.26 |
| fredross-perry: are you getting along okay? | 15:57.32 |
rayjj | chrisl: yeah, we want new glyphs, but we don't like the fact that the MD5 sums have changed :-) | 15:57.43 |
chrisl | rayjj: exactly that kind of nonsense, yes | 15:58.25 |
fredross-perry | yes I am. Wrapped up the print dialog work. Moved everything into a new repo at ghostscript.com, with mupdf and ghostpdl as submodules. | 15:58.34 |
| working through some iOS framework tutorials now. | 15:58.54 |
| I have to step out now for anther call. | 15:59.15 |
henrys | anybody else have other stuff for the meeting. I have a lot more stuff but I can just stick it on the agenda and we'll get to it as time permits. | 15:59.47 |
chrisl | henrys: I think we need to talk about hosting for our downloads at some point - the problems with silently aborted downloads don't seem to be going away | 16:00.48 |
henrys | rayjj: I don't know a list of files that don't make us look like a POS and bugs for files that do. Probably just send it to the customer and support. | 16:01.04 |
| chrisl: yup that's on the list. | 16:01.16 |
chrisl | Thanks | 16:01.20 |
henrys | tor8: jeong claim to have pretty good coverage of XFA in his product and we can have it if we want it. Do you want to look at it or reasonable probability we can't use it? | 16:03.32 |
| other meeting be back in a bit. | 16:04.14 |
rayjj | henrys: OK, for now, the JEITA files, and maybe the Altona Visual. Are there any other suggestions for inclusions. | 16:04.32 |
henrys | rayjj: that seems fine. we really do want to collect the problems you found in bugzilla too! | 16:05.20 |
chrisl | rayjj: what about the Ghent tests? They're freely available aren't they? | 16:05.22 |
rayjj | chrisl: they're mostly junk, IMHO -- not "real world" | 16:06.00 |
chrisl | rayjj: I'd agree, but being openly available means they would be tests almost anyone could repeat | 16:06.38 |
rayjj | henrys: yes, I'm pulling the list of aberrant files together for a bug. I'll start with one bug, then if we find groupings after analysis, we can close that and open separate bugs based on the nature of the problem | 16:07.38 |
| chrisl: well, any printer company will have the JEITA, and they are pretty inexpensive if not | 16:08.18 |
| chrisl: unfortunately, most printer companies will also have the QL tests and will use them | 16:08.43 |
mvrhel_laptop | henrys: just one thing. I am wrapping up a few minor issues on the gsview installer. Then I will hand it to chrisl to put on the gsview.com website as a beta version. The same installer will work for 32 or 64 bit. After that, I need to answer the color questions that one customer brought up about spot colors in gs last week and then on to SOT | 16:08.46 |
rayjj | mvrhel_laptop: I don't recall the decision on gsview. Is it going out as GPL or more restrictive (AFPL or other) | 16:10.30 |
mvrhel_laptop | the issues that I am having with the installer have to do when a user with non admin account does the install. I am being prevented from making the needed registry changes to ensure all the icons are correct and all the "open with.." options are populated. . I am thinking about just making it that you have to do the install as admin. for now. likely this is an issue that we are not signed | 16:10.47 |
rayjj | mvrhel_laptop: and does the installer have a 'click to agree' that mentions the license ? | 16:10.59 |
mvrhel_laptop | rayjj: yes. miles gave me a EULA to stick in there | 16:11.13 |
chrisl | mvrhel_laptop: I've got some stuff for handling non-admin installs that I'm going to stick in the gs installer...... | 16:11.39 |
henrys | mvrhel_laptop: will you be around in a 1/2 hour or so I had a bunch of stuff on my list for you? | 16:11.49 |
rayjj | mvrhel_laptop: great. Just checking in case it got overlooked in getting things to work :-) | 16:11.55 |
mvrhel_laptop | chrisl: great we can chat more about that later. also the gsinstaller is installing as win32 on 64 bit machines | 16:12.35 |
| the installer code for gs needs a view changes | 16:12.54 |
| I can show you what I did in gsview | 16:13.02 |
chrisl | mvrhel_laptop: if you take a look at the "nsis-updates" branch on repo, it has the revisions on it | 16:13.09 |
mvrhel_laptop | chrisl: ok | 16:13.18 |
chrisl | s/on repo/on my repo | 16:13.22 |
| mvrhel_laptop: what do mean by "installing as win32 on 64 bit"? | 16:13.50 |
mvrhel_laptop | henrys: I am supposed to run my car in for a 9am appt. when they put in the new control board for the lock they mucked up something. the trunk release opens the window! | 16:13.58 |
| chrisl. on my 64 bit machine. doing the install of ghostscript, installed the 32 bit version | 16:14.22 |
| it did not properly detect that it was a 64 bit machine | 16:14.36 |
chrisl | mvrhel_laptop: there are separate installers for 32 and 64 bit installs - you have to download the right now | 16:14.54 |
mvrhel_laptop | oh. looking at your nsis code, it looked like it was doing the decision. I compiled the nsis code in the current repos and ran it assuming it did both | 16:15.36 |
| I have a single installer for gsview that detects and installs the proper dlls in the right folder | 16:15.57 |
| we should do the same for gs | 16:16.02 |
| and only have 1 windows installer | 16:16.10 |
| but I guess | 16:16.13 |
| that some people want 32 bit on 64 | 16:16.19 |
| so maybe not | 16:16.23 |
chrisl | mvrhel_laptop: there's that, and there are complications because of the way the Windows gs build works which I haven't found a solution for yet | 16:16.56 |
mvrhel_laptop | henrys: I will be back on line about 9:15 or so after I drop the car off, if that will be OK | 16:17.07 |
henrys | mvrhel_laptop: I'll catch up with you at some point. | 16:17.31 |
mvrhel_laptop | henrys: ok | 16:17.37 |
chrisl | If I get the build issue sorted, I'd have a check box for forcing the install of 32 bit exe on a 64 bit system | 16:17.49 |
mvrhel_laptop | oh good idea | 16:18.02 |
chrisl | The single installer is on my radar (there may even be a bug for it), but it's not exactly a priority | 16:18.29 |
mvrhel_laptop | I really like the flexibility of nsis. you can do pretty much anything. including shooting your self in your foot | 16:18.40 |
chrisl | Yeh, but it also has some crazy restrictions which drive me up the wall sometimes | 16:19.10 |
mvrhel_laptop | a bit of pain to debug | 16:19.14 |
chrisl | You're not kidding!! | 16:19.31 |
Robin_Watts | mvrhel_laptop: I hope you don't mind me jumping in with that customers questions. | 16:21.35 |
mvrhel_laptop | Robin_Watts: not at all! I meant to tell you thank you | 16:21.51 |
Robin_Watts | no worries. Please do correct me if I say "Michael probably hasn't done XXXX in the .net" and you have. | 16:22.21 |
mvrhel_laptop | will do. | 16:23.07 |
| fredross-perry: do you want to stick up the installers that you have for linux and the mac on gsview.com also? not sure where we are with those | 16:30.06 |
| as a beta version | 16:30.14 |
| need to head out. back on line in about 30 | 16:34.08 |
jogux | I'll happily try any mac installer | 16:40.36 |
Robin_Watts | .quit | 16:57.09 |
henrys | kens2: I would lean to option 3. Do we know that for sure that form cache is the source of adobe speedup? (i.e.) you've studied the adobe output? | 17:00.06 |
fredross-perry | I can post some installers. But is there still an open question around whether some version of gsview will be sold? | 17:06.22 |
Robin_Watts | fredross-perry: Well, the source code is currently available for anyone to download off git.ghostscript.com, so that'd be more of a problem than some installers being up there :) | 17:09.35 |
fredross-perry | true enough | 17:10.40 |
kens | Goodnight all | 17:14.08 |
mvrhel_laptop | henrys: I am back. | 17:31.42 |
Robin_Watts | mvrhel_laptop: hey. | 17:35.28 |
| you free to talk for a mo? | 17:35.36 |
mvrhel_laptop | Robin_Watts: yes | 17:35.43 |
Robin_Watts | So, I just got another mail from the customer about .net | 17:35.53 |
mvrhel_laptop | he seems to be full of questions | 17:36.05 |
Robin_Watts | Would it be fair to say that your current .net stuff just gives enough access to the core to do what you want, rather than being a consistent exposure of the C API to .net ? | 17:37.03 |
mvrhel_laptop | Robin_Watts: that would be fair to say | 17:37.21 |
Robin_Watts | OK, so similar to what I have currently in android. | 17:37.31 |
| so, he seems to be asking if we have any plans to do a consistent API reflection into net. | 17:37.57 |
| and I would say that no, we have not got that planned. | 17:38.08 |
mvrhel_laptop | It is straight forward to add more but I would agree with you that that is not what we have planned | 17:38.28 |
Robin_Watts | How would you feel about exposing more as he wants it? | 17:38.49 |
| I suspect it'd be hard, cos even if we could get him a list of the fz_images on a page, he'd then want to be able to actually do things with those images. | 17:39.18 |
mvrhel_laptop | I don't have an issue with doing more, but I would think we would need to understand $ and what else it is going to get us. I don't know what their situation is as a customer and us doing special work | 17:41.12 |
henrys | mvrhel_laptop: this is part of what I wanted to ask about. Anything extra we need to charge them. | 17:41.20 |
mvrhel_laptop | ok | 17:41.24 |
| I was thinking the same thing | 17:41.33 |
| I had to debug his project and the issues that he had were, how should I put it.... | 17:41.58 |
| I really can't here since this is a public log | 17:42.45 |
Robin_Watts | I'll send an email back saying "no, we have no plans" and mentioning that if he wants NRE work, we can involve Miles in the discussion. | 17:43.21 |
| Seem reasonable? | 17:43.23 |
mvrhel_laptop | anyway, I think the gsview project provides a nice demo of what can be done. Robin_Watts yes sounds good | 17:43.38 |
| I would like to add more capabilities as time moves on but I think for now it is done in terms of capabilities. otherwise my other pots are going to boil over | 17:44.35 |
Robin_Watts | yeah, I hear that. | 17:44.45 |
henrys | mvrhel_laptop: I also wanted to talk to you about nsis but you already discussed it. How much time are you anticipating with the color customer? | 17:45.47 |
mvrhel_laptop | henrys: maybe 2 days at most | 17:46.21 |
| I might be able to figure it out doday I hope | 17:46.39 |
| today | 17:46.41 |
| henrys: you have me nervous. is there some pressing issue on the horizon? | 17:48.35 |
| other than what I know | 17:48.45 |
henrys | mvrhel_laptop: nope just wanted to ask about the .NET and nsis business. | 17:49.12 |
mvrhel_laptop | ok good deal. | 17:49.28 |
mvrhel_laptop_ | hmm network at this deli (as I wait for the car) is not so good | 17:51.24 |
| breakfast sandwich was good though | 17:51.37 |
Robin_Watts | helen is teaching small child downstairs. parent of small child is waiting for lesson to be over with smaller child. smaller child is a noisy little oik prone to screaming. | 17:56.43 |
| Must. not. touch. shotgun. | 17:56.53 |
mvrhel_laptop | hehe | 18:15.12 |
| chrisl; you still there? | 18:15.17 |
chrisl | mvrhel_laptop: for a very short time, yes | 18:16.35 |
mvrhel_laptop | ok I am going to stick the installer in the gsview directory on casper | 18:16.53 |
| if you want to grab it later and put it on the web stie | 18:17.04 |
| site | 18:17.06 |
| I just need to stop and say enough is enough with this | 18:17.29 |
| and take the brunt of comments.... | 18:17.39 |
chrisl | Sure, do we want to get everyone (using Windows) in the company to try it, before making it public? | 18:17.53 |
mvrhel_laptop | that would be a good idea | 18:18.00 |
Robin_Watts | Gimme a location and I'll give it a whirl. | 18:18.22 |
mvrhel_laptop | ok hold oon | 18:18.28 |
| on | 18:18.30 |
| pushing it up. network is slow here... | 18:21.14 |
| installer is good sized since it is for both win32 and x64. Another reason to do 2 different ones | 18:22.20 |
| ok | 18:22.31 |
| ok its in ~mvrhel/gsview/setup.exe | 18:22.36 |
| so I would recommend you are in your admin account when you install | 18:22.53 |
| There is something going on with it not being signed and UAC stuff that is causing issues in getting the open with options populated in the registry otherwise | 18:23.47 |
chrisl | That's on Win8? | 18:24.21 |
mvrhel_laptop | I installed on a 32 bit win 7 machine last night | 18:24.40 |
| but from a user account | 18:24.45 |
| and it had the same issues as in win8 | 18:24.52 |
| from a user account | 18:24.55 |
| I am going to try the admin account on the win 7 machine when I get back home | 18:25.06 |
| to see if the behavior is the same | 18:25.16 |
chrisl | As a non-admin user you can basically only install for the current user only, because you can only write to HKEY_LOCAL_USER, not HKEY_LOCAL_MACHINE | 18:25.39 |
| Ugh, too many "onlys".... | 18:25.59 |
mvrhel_laptop | yes the HKEY_LOCAL_MACHINE is the issue | 18:26.04 |
| but I am forcing run as admin | 18:26.13 |
| and it still does not work | 18:26.19 |
| lots of discussion about this in https://msdn.microsoft.com/en-us/library/windows/desktop/ms724475%28v=vs.85%29.aspx | 18:27.00 |
chrisl | Okay, nsis has some features to switch between LOCAL_MACHINE and LOCAL_USER - that's some of what my changes to the gs installer do. But it is a pain | 18:27.08 |
mvrhel_laptop | chrisl: I need to look at what you have done. also I need to get this stuff committed | 18:27.44 |
chrisl | Erm, we really should be sending out a beta with uncommitted code.... even for the installer | 18:28.19 |
mvrhel_laptop | well it will be committed shortly | 18:29.13 |
chrisl | mvrhel_laptop: it should really be tagged as a release | 18:29.36 |
mvrhel_laptop | ok. well lets beat on it internally first | 18:29.56 |
| at some point we need to peel the project out of the mupdf repos | 18:30.35 |
| tor8 had started that | 18:30.41 |
| perhaps this would be a good time to do that | 18:30.47 |
chrisl | mvrhel_laptop: so, (first complaint!) can you call it something other than "setup.exe", like gsvsetup.exe or something? | 18:30.57 |
mvrhel_laptop | yes | 18:31.06 |
| let me do that now | 18:31.26 |
Robin_Watts | gsview_setup.exe. These are not 8+3 days :) | 18:31.29 |
mvrhel_laptop | even better | 18:31.43 |
chrisl | Robin_Watts: Sacrilege! | 18:32.07 |
mvrhel_laptop | ok fixed that | 18:32.45 |
chrisl | mvrhel_laptop: I'll have a proper play with it tomorrow, but I'll hold off putting up for public consumption for a few days. | 18:32.58 |
marcosw | Robin_Watts: "gsview setup.exe" (if we've moved past 8.3 we can have spaces too). | 18:33.05 |
mvrhel_laptop | yes. good idea | 18:33.07 |
Robin_Watts | aargh. No. spaces are a bad thing. | 18:33.30 |
mvrhel_laptop | yes no spaces | 18:33.36 |
chrisl | marcosw: I *will* hit you - spaces... hmph! | 18:33.42 |
mvrhel_laptop | the yes good idea was for chrisl about holding off | 18:34.19 |
chrisl | Right, have to go..... goodnight all! | 18:35.17 |
mvrhel_laptop | night chrisl | 18:35.24 |
| Robin_Watts: I have the installer changes in my repos on casper. If you can have a look | 18:42.20 |
Robin_Watts | Is platform/windows/gsview/LICENSE a text file for human consumption? Or does it get read in and used by nsis? | 18:43.30 |
mvrhel_laptop | read in and used | 18:43.38 |
Robin_Watts | If the former, wordwrapping it would be good. | 18:43.39 |
| ah, ok. | 18:43.42 |
mvrhel_laptop | I spent a bit of time spacing it so that it was readable in nsis | 18:43.56 |
| I should move it to the nsis folder | 18:44.21 |
Robin_Watts | gsview_setup_$VERSION ? | 18:44.26 |
| Does that work? | 18:44.27 |
mvrhel_laptop | that would work | 18:44.36 |
| good diea | 18:44.49 |
| idea | 18:44.51 |
Robin_Watts | ok, I can't claim to understand much of it, but it seems plausible :) | 18:45.52 |
mvrhel_laptop | :) ok. you found a couple things. I am going to move the license file to the nsis folder and I will make is to that installer name has the version number | 18:46.24 |
| then I will push to golden | 18:46.31 |
| bbiaw. need to check on car | 19:02.06 |
kens | henrys I missed your question at 17::, sorry about that. | 19:54.50 |
fredross-perry | sview install question: do I need to present a license to end users? What license? thanks. | 19:55.48 |
kens | Studying the Adobe *output* won't, unfortunately, tell us anything about how its handling the input. However, given that the input is nearly 5000 pages, with 3 different forms, and Distiller runs in 16 seconds, it clearly cannot be executing the form PaintProc that many times | 19:55.55 |
| Also, that is where the bulk of the time goes even for pdfwrite. Actually determining duplicates is quite quick, because we hash the streams as we go. SO if we did not execute the form PaintProc thousands of times, we would also not need to write the acumulated form stream and then throw it away when we discover a duplicate. | 19:57.33 |
| Obviously I don't know for sure without trying it what this would win us in performance terms for pdfwrite, but I'm willing to bet its a *lot*. | 19:58.02 |
| And now, back to TV...... | 19:58.15 |
Robin_Watts | fredross-perry: The same license as michael does I guess... hold on... | 19:59.09 |
| http://git.ghostscript.com/?p=user/mvrhel/mupdf.git;a=blob;f=platform/windows/gsview/LICENSE;h=4fd64b191152d9661e14e9f734a225a3ee51efb6;hb=647210a7a8bdad10948c07415c9701e032016650 | 19:59.33 |
fredross-perry | what URL do things in public_html get? | 22:45.53 |
| OK, I see it now. | 22:47.24 |
| I put gsview installers for mac and linux in http://ghostscript.com/~fred/gsview - have fun | 23:01.23 |
| Forward 1 day (to 2015/02/04)>>> | |