| <<<Back 1 day (to 2014/03/19) | 2014/03/20 |
Robin_Watts | zeniko: (For the logs) yes. | 00:10.19 |
Jogux | I think the ATS VMs possibly aren't going to get us as far forward as we hoped; for not entirely clear reasons the VM images were setup from scratch rather than imaged from the working servers, and it seems like the setup may not have been completed :( | 00:10.42 |
Robin_Watts | Images might change subtley because of the l2factor that they are decoded at. What sort of changes are you seeing? | 00:11.48 |
pedro | Jogux: do you remember if the ATS server disks were sold off as part of the kincaid auctions? | 00:37.27 |
| I guess we can get the active build commands from the ATS server config, but build env setup may be more of a pain :( | 00:39.01 |
| at least the release tracking spreadsheet seems reasonably up-to-date and complete | 00:42.49 |
pedro | offs - snowblind from ploughing through office xml files. Tomorrow is another day... | 01:18.29 |
mvrhel_laptop | blah. I see that the stream API has been changed in mupdf... | 06:04.01 |
| will try to fix winRT stuff in the morning | 06:04.21 |
Robin_Watts | mvrhel_laptop: oops, yes, sorry | 09:10.53 |
paulgardiner | I have to go out for a while baiw | 09:45.32 |
Robin_Watts | new git conversion phase 1 complete. Doing the git fast-import now. | 09:48.01 |
| Jogux: So, this git filter branch stuff to remove "Summary:" from every commit... | 10:04.56 |
| How do we do that? | 10:05.04 |
Jogux | something like git filter-branch --msg-filter 'sed s/^Summary:\n//' | 10:06.37 |
Robin_Watts | Will that run on every commit? OR just every commit on the branch ? | 10:08.03 |
Jogux | I think you need a --all on the end | 10:09.25 |
| you may want to make a copy of the repository first :-) | 10:09.36 |
| oh, possibly -- --all actually | 10:10.40 |
Robin_Watts | Normally -- means "and what follows are pathnames, not tags" | 10:12.06 |
| so -- --all would mean "all files" surely? | 10:12.18 |
Jogux | the -- in this case separates filter-branch options from revision options, apparently. | 10:13.44 |
Jogux | is reading man git-filter-branch | 10:13.57 |
| I'm sure there's a smarter way to do that sed too. | 10:14.14 |
Robin_Watts | Ass. I think peeved just fell off the net :( | 10:15.10 |
| but http://mm0hai.net/blog/2011/03/10/rewriting-git-commit-message-history.html suggests that you were right. | 10:15.30 |
kens | Looks like Ray's office is offline or powered down, both peeves and peeved are down | 10:19.10 |
Robin_Watts | might be back now. | 10:33.34 |
kens | Seems to be back on teh dashboard | 10:33.43 |
Robin_Watts | running now. This will take a while. going for a run. | 10:40.47 |
tor8 | Jogux: sed '/^Summary:$/d' | 11:07.19 |
Jogux | yeah, that sounds better, ta | 11:09.09 |
tor8 | I don't remember if filter-branch runs the edited commits through 'git stripspace' like normal commits do | 11:09.41 |
| if not, we may have to do that to so as not to get a blank line the first line of the commit | 11:09.59 |
| s/to/too/2 | 11:10.12 |
Jogux | nods. | 11:10.18 |
binarious | Hi, is there a way to reverse the whole page order in a pdf document? | 11:21.05 |
kens | You could edit the pages tree | 11:21.52 |
binarious | I was hoping for a simple ghostscript command | 11:22.50 |
kens | There is no 'simple' command, also as I keep telling people, when you use pdfwrite on a PDF file you don't just 'modify' the existing file, you create a totally new one. Only the appearance is the same | 11:23.44 |
| Currently the PDF interpreter is difficult to script, so its not easy to render pages out of order. | 11:25.14 |
tor8 | binarious: you can use mupdf: mutool clean -o rev.pdf input.pdf 5 4 3 2 1 | 11:25.36 |
| to select a subset of pages and their ordering in the output | 11:25.46 |
kens | Ah, well that was going to be my next suggestion | 11:25.50 |
binarious | That's fine. I'd like to create a new reversed file :p. Just like combining multiple pdfs into one now I'd like to reverse them. | 11:25.54 |
kens | binarious : using MuPDF will be a much simpler task | 11:26.13 |
| possibly also pdftk but I'm not sure about that one | 11:26.45 |
| You *can* do it with Ghostscript, but its non-trivial | 11:27.05 |
| I really must rejig the PDF interpreter so its not so difficult | 11:27.17 |
binarious | tor8: well I don't want to combine the files reversed. I'd like to reverse the whole page order of a merged pdf, because the seperate pdf files have the "wrong" page order, too. | 11:27.27 |
kens | I'm not sure I follow that | 11:28.00 |
| tor's command line will reverse a PDF file, I thought that's what you wanted | 11:28.13 |
| Of course, you would need to know the number of pages, but that's trivial | 11:28.38 |
| Looks like it isn't too difficult with GS actually, since we don't want to do N-up or anything complex with teh graphics state | 11:30.07 |
tor8 | binarious: if you have the merged pdf, the command line I gave will reverse the internal page order | 11:35.22 |
| Robin_Watts: damn. we've broken 'mutool info' :( | 11:36.08 |
imran_591 | i need some help regarding mupdf | 11:36.27 |
| can anybody help ? | 11:37.06 |
tor8 | imran_591: ask. don't ask to ask. | 11:37.17 |
| nobody knows if they can help if you don't first say what you want to know | 11:37.47 |
imran_591 | if(core.hasOutline()) method returns null for some pdf and for some it returns true. am loading pdf from local | 11:38.48 |
| in MuPdfActivity.class | 11:39.00 |
kens | That isn't a question, its a statement | 11:40.51 |
imran_591 | Whatever Kens ? thats a doubt | 11:41.53 |
kens | I'm sorry I don;t understand, you've made a statement, you haven't asked a question. | 11:42.15 |
tor8 | an if statement doesn't return anything... core.hasOutline() returns true or false, never null. | 11:42.43 |
imran_591 | whatever please let me finish it | 11:43.27 |
binarious | tor8: Ah okay, but that would require to list the pages? n n-1 5 4 3 2 1? | 11:43.29 |
kens | Yes | 11:43.38 |
tor8 | binarious: yeah, I think we lost the automatic reversion at some point. it used to suffice to list the pages in reverse order :( | 11:44.18 |
| I mean ranges in reverse order | 11:44.29 |
| like mutool clean 5-1 | 11:44.50 |
binarious | tor8: well but I have about 200k pages. I think a command line wont take that much input :( | 11:45.04 |
kens | That's... a lot of pages.... | 11:45.16 |
imran_591 | i want to know why it returns false, when many pdf returns true.some pdf false. | 11:45.19 |
tor8 | I might add that back in, but for now just use $(seq 200 -1 1) | 11:45.25 |
| binarious: oh. that makes it more urgent :) | 11:45.32 |
kens | imran_591 : then you should ask that question :-) | 11:45.34 |
imran_591 | kens :) sorry | 11:45.44 |
kens | My guess would be that some PDF's have outlines, others don't | 11:45.49 |
imran_591 | outlines means ? | 11:45.57 |
binarious | Yep, that raises a bit more problems :) | 11:45.59 |
tor8 | imran_591: hasOutline tests if the PDF has an outline (the bookmark list/table of contents thing) | 11:46.04 |
kens | imran_591 : Outlines are defined in the PDF reference | 11:46.11 |
imran_591 | i cant understand | 11:46.45 |
| :( | 11:46.47 |
binarious | brb thanks so far | 11:46.58 |
tor8 | binarious: well, I'm doing documentation today... so I may have a quick look | 11:47.09 |
kens | A PDF file may *optionally* have outlines (bookmarks) but its not a requirement. THe test tells you whether a file does or not | 11:47.19 |
tor8 | it might be a tiny one line patch | 11:47.19 |
imran_591 | ya ya i got you.Let me check it | 11:47.40 |
tor8 | binarious: okay, I have a patch for you. | 12:02.31 |
| binarious: http://git.ghostscript.com/?p=user/tor/mupdf.git;a=commitdiff;h=0cb41a5e1cdedc7e48a897292bbfaf85f30ad7e0 | 12:04.39 |
| binarious: with that patch you can run: mutool clean -o reverse.pdf input.pdf -1 | 12:05.28 |
| the missing number before the - implies the last page | 12:05.49 |
Robin_Watts | tor8: We've broken all windows builds too. | 12:08.06 |
| tor8: See the patches on robin/master | 12:08.12 |
tor8 | Robin_Watts: I've taken on the printf one | 12:08.30 |
| and okayed the one after that | 12:08.38 |
| and asked if you wanted the third one in before release | 12:08.46 |
| but I think those messages scrolled past your eyes unseen :) | 12:09.08 |
Robin_Watts | tor8: oh, sorry, when was that? | 12:10.30 |
tor8 | last night | 12:10.36 |
| just before I went to bed | 12:10.44 |
Robin_Watts | ah! | 12:10.50 |
| sorry. | 12:10.52 |
| There are 4 reviews on there. | 12:11.18 |
| no, ignore me. | 12:11.26 |
| yes, I'd like to get the third one in before the release. | 12:11.35 |
tor8 | Robin_Watts: there's a fix on tor/master to respect negative ranges in mutool clean (like mudraw) | 12:11.38 |
Jogux | if only we had a review system! :-) | 12:11.49 |
tor8 | Jogux: I fear what you see is as good as it gets around here ;) | 12:12.29 |
stars | Hi all | 12:13.29 |
Jogux | tor8 : I'm partly being silly as a web based 'review system' is one of the chunks of code obtained from picsel :-) | 12:13.43 |
tor8 | Robin_Watts: LGTM (have you tested them?) | 12:13.44 |
Jogux | but, tbh, for this stuff I could see gitlab/stash/whatever style pull-requests helping. (the picsel system was very tied to cvs etc) | 12:14.20 |
tor8 | Jogux: I've seen various other web base review systems around. they all seem rather awkward compared to IRC when there's only a handful of developers who mostly work in the same timezone | 12:14.36 |
Robin_Watts | tor8: I have. | 12:14.46 |
tor8 | Robin_Watts: push away then :) | 12:14.55 |
| might as well break everything before we do the release freeze | 12:15.03 |
| I was hoping to smash in the mujs stuff today as well | 12:15.16 |
| now that the name seems to be settled | 12:15.21 |
Robin_Watts | Jogu: We use the web view on git.ghostscript.com for reviews. Comments etc happen on here. | 12:15.33 |
| An automated system would be nice, but... this suffices. | 12:15.56 |
Jogux | nods. | 12:16.04 |
binarious | tor8: awesome thank you! So I can check out the latest source of mupdf then apply your patch and compile? | 12:23.52 |
tor8 | binarious: yes. (or check out my version directly) | 12:24.27 |
| or wait until Robin_Watts givs the nod to the commit and it goes into the master repository | 12:24.55 |
binarious | that'd be my fav solution :) | 12:25.25 |
tor8 | hi stars | 12:25.30 |
Robin_Watts | tor8: nod. | 12:28.53 |
stars | how is_list_item and is_list_item_ui are intended to be used.... by checking each character parsing or is it designed to be parsed by line by line in stext-paragraph.c | 12:29.34 |
Robin_Watts | stars: What version of mupdf are you looking at? | 12:30.48 |
tor8 | binarious: okay, it's landed on the master repo | 12:31.24 |
stars | Robin_Watts, 1.3 | 12:31.55 |
| trunk infact | 12:32.00 |
Robin_Watts | stars: I have an "is_list_entry", but neither of the functions you mention. | 12:32.03 |
binarious | tor8: Robin_Watts: thanks for the fast reaction! | 12:32.14 |
stars | sorry | 12:32.24 |
| I am talking about is_list_entry and is_list_entry_ui | 12:32.39 |
Robin_Watts | I have no is_list_entry_ui | 12:33.08 |
| is_list_entry is a function called by fz_analyze_text | 12:33.36 |
| It looks at a given span on a line to try to guess if it's a list entry or not. | 12:33.56 |
| All this code is hairy heuristics. | 12:34.08 |
stars | Robin_Watts, now if I were to include <li> tags should I be doing it in the stext-output.c? | 12:36.46 |
| if I try using is_list_entry to speculate a character ... in case I encounter more than one span in a line, it is going to add multiple <li>tags | 12:37.43 |
| or I am completely using it the wrong way! | 12:37.58 |
Robin_Watts | stars: The idea of fz_analyze_text is that it takes the lines/spans/chars structures produced by the raw extraction from the PDF page, and attempts to make guesses about how they were laid out. | 12:40.05 |
| Specifically it does things like look for hanging indents, common lineheights etc. | 12:40.40 |
| That hopefully allows it to spot paragraphs. | 12:40.51 |
| It basically produces the same lines/spans/chars structures, but with the data rearranged slightly and with some annotations (like lineheights, column start/end positions) added. | 12:41.35 |
| The code in stext-output.c then takes these new structures and outputs them as text. | 12:42.04 |
| or html. | 12:42.12 |
| So in general 'analysis' stuff should go in stext-paragraph.c, and html specifics should go in stext-output.c | 12:42.51 |
Robin_Watts | pops to shop. back in 5 mins. | 12:43.44 |
stars | Robin_Watts, what I am trying to do is to get the li tags working with fz_print_style_begin and fz_print_style_end | 12:43.55 |
| of stext-output.c | 12:44.14 |
| I am clueless to get it as a flag. the way get it for <sup> or <sub> | 12:46.40 |
Robin_Watts | stars: Part of the 'style' record that we keep is whether it's a superscript or a subscript. | 13:05.49 |
| hence in print_style_begin we can send sub or sup. | 13:06.05 |
| being a list item or not is not a member of the style record. | 13:06.25 |
| hence you can't do it as part of fz_print_style_begin. | 13:06.58 |
stars | Robin_Watts, so is there any other suggestion that you can give me in order to resolve this...please? | 13:19.07 |
| is it something to implement with the span ? | 13:19.40 |
Robin_Watts | stars: Sorry, I'm really not clear on exactly what you're trying to achieve. | 13:20.00 |
stars | Robin_Watts, I am only trying to add li tags while it is a list in the html extracted | 13:20.28 |
Robin_Watts | So, in stext-paragraph.c you need to modify the code to spot that it's in a list, and to store that fact in the line or span record. | 13:22.08 |
| That will mean updating the line or span structure I suspect. | 13:22.21 |
| Then in stext-output.c you can look at that code and output as appropriate. | 13:22.45 |
tor8 | Robin_Watts: paulgardiner: 4 commits on tor/master for review when you have a minute | 13:29.00 |
Robin_Watts | seriously, our random number generator is return 0.4 ? | 13:30.12 |
tor8 | Robin_Watts: for the cluster tests, yes | 13:30.21 |
Robin_Watts | Ah! | 13:30.26 |
binarious | Do you know if I can sort the batch processing input for gs? For example: "gs -q -dNOPAUSE -dBATCH -sDEVICE=pdfwrite -sOutputFile=merged.pdf *.pdf" would process the following files: "1.pdf, 2.pdf, 3.pdf, 10.pdf" as follows: 1.pdf, 10.pdf, 2.pdf, 3.pdf". | 13:31.02 |
kens | Thats a shell problem | 13:31.22 |
tor8 | binarious: $(ls *.pdf | sort) | 13:31.33 |
| and pray that you have no spaces in the filenames | 13:31.45 |
binarious | As the parameter for *.pdf? | 13:31.50 |
Robin_Watts | tor8: HAVE_X11 ?= yes sets HAVE_X11 to yes, unless it's already set. | 13:31.53 |
tor8 | instead of *.pdf | 13:32.01 |
binarious | thanks again :) | 13:32.09 |
Robin_Watts | HAVE_X11 := yes always sets it, unless it was specified on the make invocation. | 13:32.25 |
tor8 | Robin_Watts: hm, right... | 13:32.53 |
Robin_Watts | So AIUI, previously: "HAVE_X11=no make" would have suppressed X11. | 13:33.06 |
| but with the new code it won't. | 13:33.10 |
tor8 | Robin_Watts: yeah :( | 13:33.27 |
Robin_Watts | First 2 look fine, 3rd one we just spoke about. looking at 4 now. | 13:34.39 |
norbertj | henrys_: the gap in the jpegimage is caused by passing the data into the decoder in small parts (because of the size of the inputbuffer (1<<9, 1<<13 in plmain.c line 313). If I make the buffer large enough (1<<15, for the testfile), then the gap disappears. | 13:36.17 |
Robin_Watts | tor8: I am confused by the 4th one. | 13:36.54 |
tor8 | Robin_Watts: too much in one commit? | 13:37.28 |
| or the makefile stuff? | 13:37.33 |
Robin_Watts | Previously mudraw was built with $(MUPDF_LIB) $(MUPDF_JS_NONE_LIB) $(THIRDLIBS) | 13:37.38 |
| and now you've dropped $(MUPDF_JS_NONE_LIB) | 13:37.58 |
| how does that work? | 13:38.05 |
tor8 | I moved the JS lib into the main one, so we always build everything against the chosen javascript lib | 13:38.24 |
Robin_Watts | or do we only build one MUPDF_LIB now ? | 13:38.27 |
| right. | 13:38.31 |
tor8 | for mudraw that should get eliminated at link time | 13:38.38 |
Robin_Watts | We can hope :) | 13:38.53 |
tor8 | hm. better make sure, I guess! | 13:39.22 |
| but I need to fix the makefile problem you mentioned | 13:39.32 |
| I didn't consider the HAVE_X11 = no case | 13:39.43 |
| just the presence of the variable | 13:39.46 |
paulgardiner | pedro_: looks like the windows build needs openssl. Do you know what version we should use? Or at least which worked for you? | 13:40.58 |
henrys_ | norbertj:that should be okay design wise. I'm pretty sure ghostscript uses the jpeg decoder with small buffers so something in pximage.c... | 13:41.35 |
Robin_Watts | paulgardiner: epage/platform/tgv/apps/windows-alien-vs2005/openssl ? | 13:43.54 |
paulgardiner | 2010 | 13:44.10 |
Robin_Watts | epage/third-party/tropicssl/include/tropicssl/openssl.h ? | 13:45.19 |
tor8 | Robin_Watts: oh bollocks. the js library got pulled into mudraw :( | 13:45.44 |
Robin_Watts | tor8: Doesn't entirely surprise me :( | 13:46.01 |
| paulgardiner: copy the one from epage/platform/tgv/apps/windows-alien-vs2012/openssl ? | 13:46.45 |
stars | Robin_Watts, Sure I will give it a try | 13:46.52 |
| thanks for all the help! | 13:46.58 |
tor8 | Robin_Watts: yeah, pdf_init_document initialises the js | 13:47.21 |
binarious | tor8: btw pdftk does that, too. (pdftk merged.pdf cat end-1 output out.pdf) | 13:47.39 |
paulgardiner | robin_watts could I guess, but my thinking was that pedro_ just had this build work, so I just need to copy his setup | 13:47.45 |
binarious | (the reverse thing) | 13:47.49 |
Robin_Watts | tor8: 4th one looks fine to me other than that, but you probably need paulgardiner to look to get a proper review. | 13:49.23 |
paulgardiner | tor8, robin_watts will look now | 13:49.51 |
tor8 | Robin_Watts: well, we're hosed anyway... some other changed has pulled all the embedded fonts into mutool as well | 13:50.17 |
| I suspect decouple_t3_fonts | 13:50.27 |
norbertj | henrys_: I checked with the -Zw -ZI to see what data is consumed from the input, and it seems that is all ok (i.e. the px_process handling of leftoverdata+newdata seems ok). | 13:50.33 |
tor8 | Robin_Watts: makefile fixes in a new commit on top of those 4 | 13:51.05 |
| I'll rebase that one down before pushing, but to see the changes more easily I made it a separate commit for now | 13:51.19 |
henrys_ | norbertj: oh so you think it is the decoder? | 13:51.43 |
Robin_Watts | tor8: looks better to me. | 13:53.01 |
tor8 | Robin_Watts: paulgardiner: I'm thinking we should make the pdf_jsimp calls based on function pointers and maybe use lazy initialization | 13:54.33 |
| or we could give up trying to have a small binary for mutool... | 13:54.59 |
| but I fear we have spaghetti dependencies that've landed us in this mess again :( | 13:55.18 |
Robin_Watts | tor8: Can we move the js init from pdf_document_init to the one that's used only for run ? | 13:55.29 |
tor8 | Robin_Watts: that's what I was thinking | 13:55.39 |
| and then the pdf_drop_js call could be a function pointer | 13:55.50 |
pedro_ | paulgardiner: I just used the most up-to-date openssl windows build (still has to be in c:\openssl I think) | 13:55.54 |
tor8 | likewise with the t3 decoupling | 13:55.58 |
paulgardiner | tor8: so with "%*s", in addition to controlling the printed width, the count will prevent running off the end of a non-delimited string? (I should know this, but ...) | 13:56.03 |
tor8 | and whatever else has dragged in the graphics library into the mutool | 13:56.08 |
pedro_ | paulgardiner: http://slproweb.com/products/Win32OpenSSL.html | 13:56.38 |
Robin_Watts | pedro_: Probably we should make a place on casper to store all these required build components. | 13:57.11 |
paulgardiner | pedro_: ta | 13:57.35 |
tor8 | paulgardiner: yes, but that WIP commit shouldn't make it onto master | 13:57.37 |
paulgardiner | pedro_: Any other installs I'm likely to need? | 13:57.54 |
norbertj | henrys_: I think that the bufferdata passed in correctly reflects what is consumed (i.e. par->source is correctly updated by read_bitmap/jpegdecoder), So pxparse.c correclty passed new bufferdata. | 13:57.58 |
pedro_ | Robin_Watts> yup, probably worthwhile. We used to keep build config details on the wiki | 13:58.10 |
paulgardiner | We should get the wiki up and running again. I might look at that in a moment if no one else is on it | 13:59.15 |
Robin_Watts | paulgardiner: We should. | 13:59.30 |
pedro_ | paulgardiner: can't think of anything offhand (although I do already have directx sdk and platform SDK installed here) | 13:59.31 |
norbertj | henrys_: so indeed I think the jpegdecoder gets a bit confused. | 13:59.57 |
paulgardiner | pedro_: okay. I don't, but that's good because I can establish if they are required | 14:00.18 |
pedro_ | paulgardiner: we shouldn't actually need the DX9 stuff for the vs2010 build iirc | 14:00.34 |
henrys_ | norbertj: or the image cursor bookkeeping in pxl | 14:01.30 |
tor8 | paulgardiner: do you know where mutool might have a dependency on pdf-form from? | 14:01.43 |
Robin_Watts | tor8: #if 0 ... #endif around pdf-form.c and try a link? | 14:02.18 |
paulgardiner | tor8: a complete guess would be something to do with measuring text | 14:02.36 |
norbertj | henrys: that would be the par->source stuff? | 14:03.06 |
tor8 | paulgardiner: Robin_Watts: subtle... fz_javascript_supported | 14:03.54 |
paulgardiner | tor8: so which commits should I be looking at? | 14:03.58 |
tor8 | paulgardiner: all but that one :) | 14:04.17 |
| Robin_Watts: right, so the second culprit is pdf_load_(link_)annots and pdf_free_annot in pdf-page, the link dests and actions in pdf-outline and annot iteration function pointer setting in pdf-xref | 14:06.16 |
paulgardiner | tor8: the allocator one: is that so mujs has the context when comes to free memory? | 14:06.33 |
tor8 | paulgardiner: sorry, where are you looking? | 14:07.39 |
| (or rather, I don't understand the question) | 14:07.55 |
henrys | norbertj: no I was thinking the image data was being fed to the graphics library wrong when we ran out of data, but that can't be it. | 14:08.23 |
paulgardiner | gah! sorry. Looking at the wrong branch. It's working on this cramped laptop screen. That's my excuse | 14:09.05 |
norbertj | henrys: you mean the pl_image_data()? That does exactly render what the read_bitmap() produced. | 14:09.38 |
henrys | yes | 14:09.59 |
| norbertj: actually if you add -ZB to the switches and compare results it should isolate the issue - the -ZB will dump the bitmap and positions. | 14:10.51 |
| so -ZIBW | 14:11.15 |
| get rid of banding with -dMaxBitmap=1G so that doesn't confuse matters. | 14:11.43 |
| new fonts ordered - yeah! | 14:16.47 |
| Robin_Watts, kens, mvrhel_laptop was there a review for hin-tak's commit that I missed? | 14:18.02 |
kens | not that I saw | 14:18.23 |
Robin_Watts | henrys: not by me. | 14:18.39 |
paulgardiner | What's the significance of the block structure? To aid reading? | 14:18.42 |
henrys | we all agree someone on the staff should review non staff stuff yes? | 14:18.58 |
Robin_Watts | henrys: In this instance, looking at the commit, it's just typos in comments. | 14:19.47 |
paulgardiner | tor8: ^ | 14:20.21 |
kens | it didn't look controversial at all | 14:20.41 |
Robin_Watts | but, as you know, I'm in favour of every commit (non-staff or otherwise) having at least one other pair of eyes glance at it before it goes in. | 14:20.58 |
| it costs very little and saves many silly slips. | 14:21.22 |
henrys | kens: true but I think everything should be checked from the outside - we could be in freeze or some other state | 14:21.48 |
kens | Yes, not disagreeing | 14:21.58 |
Robin_Watts | likewise. | 14:22.06 |
Jogux | it may be sane to stop any non-staff member even pushing to 'master'? (code could be pushed by the staff reviewer instead?) | 14:22.44 |
tor8 | paulgardiner: if you mean the block structure in in jsimp_addmethod, then yes, just for readability | 14:22.53 |
| it mirrors the value stack depth/scoping | 14:23.04 |
Robin_Watts | Jogux: Yeah, the current permissions thing predates our use of git. | 14:23.29 |
tor8 | henrys: so which ones are we getting, and under what license? | 14:23.50 |
henrys | tor8:gpl for the 35 only | 14:24.32 |
tor8 | full cyrillic complement for all 35? | 14:24.57 |
henrys | tor8: yes | 14:25.05 |
tor8 | henrys: that's nice! | 14:25.06 |
| any ETA for when we'll have them? | 14:25.16 |
paulgardiner | tor8: I think I'm going to have to mostly just trust you on this one. I'd need a more detailed understanding of the workings of mujs to review this in detail (although I have to admit it is something I'd like to look at when I get a chance) | 14:25.30 |
tor8 | paulgardiner: I'd be happy to walk you through it when you get a chance :) | 14:25.56 |
| paulgardiner: I'm going to split pdf-annot.c into two files | 14:26.09 |
| one for loading, and the other for all editing | 14:26.14 |
paulgardiner | great. that would be good | 14:26.16 |
tor8 | the editing stuff depends on the graphics library, but loading the links for mutool use shouldn't need to pull that in | 14:26.39 |
| any suggestions for names? | 14:26.41 |
paulgardiner | My comment works as an answer to both yours | 14:26.44 |
henrys | tor8: they are starting with the 35 - 3 months | 14:26.54 |
tor8 | henrys: aight. so not in time for mupdf 1.4 then :) | 14:27.06 |
Robin_Watts | Jogu: I'm not convinced that that git rewrite did anything :( | 14:28.34 |
| yeah, master still points to the same checksum. | 14:29.37 |
tor8 | Robin_Watts: check the refs/ tree, if you have a refs/original/heads or something like that | 14:31.36 |
Robin_Watts | tor8: The size of my 2 git repos (backup and the one I just ran on) is identical. | 14:31.55 |
| I have no refs/originals | 14:32.24 |
| I have no refs/original | 14:32.27 |
tor8 | Robin_Watts: hm, git-filter-branch should always create some 'original' set of refs... | 14:33.13 |
| at least that was the case when I did the ghostpdl git conversion | 14:33.32 |
Robin_Watts | tor8: Is there some way to only rewrite the last 100 entries or something? | 14:34.37 |
tor8 | Robin_Watts: yes (but I forget the exact syntax) | 14:34.56 |
| which command line did you use? | 14:35.15 |
Robin_Watts | ok, I might see what went wrong. | 14:36.16 |
| Ok, I was using: git filter-branch --tag-name-filter cat --msg-filter 'sed s/^Summary:\n//' | 14:36.46 |
| Without the \n it works fine. | 14:36.53 |
tor8 | Robin_Watts: yeah, the \n won't match | 14:37.06 |
| use '/^Summary:$/d' instead | 14:37.14 |
| 'sed /^Summary:$/d' | 14:37.43 |
Robin_Watts | OK, running again now. | 14:40.07 |
norbertj | henrys: if I run with ZIBW and buffers 512 resp 32k I see that the small is missing scan (y=7) i.e. the largescan[8] == smallscan[7], largescan[9] = smallscan[8] | 14:42.15 |
| So scans (counting from 0 to 31) 7, 15, 23, 31 are missing when using small-buffersize. | 14:44.52 |
spanners | hello | 14:48.09 |
Robin_Watts | spanners. | 14:48.46 |
| Everyone meet spanners. He wrote the Picsel Font engine. | 14:49.00 |
spanners | s/wrote/fixed/. Hello! | 14:49.10 |
norbertj | henrys: and when using buffersize of 8k (release-default), then scans 15 and 23 are missing. | 14:49.15 |
kens | Hi | 14:49.21 |
ghostbot | hey, kens | 14:49.21 |
kens | LOL | 14:49.26 |
chrisl | spanners: Hi, welcome to the asylum..... :-) | 14:50.30 |
kens | He does fonts, he's beein sectioned for ages I should think | 14:51.09 |
norbertj | henrys: note I use the following: pcl6 -sDEVICE=display -ZIBW -r300 bug.xl 2>&1 | tee logfile, And then I filter on image and row. | 14:51.34 |
spanners | I used to do fonts. Now I do anti-tamper software for this lot: http://metaforic.com/ | 14:51.47 |
chrisl | Yes, sorry, I should have said "to *this* asylum".... | 14:52.04 |
tor8 | hi spanners! | 14:53.43 |
henrys | welcome spanners | 14:56.39 |
pedro_ | hi spans | 15:07.37 |
paulgardiner | hi spans | 15:09.48 |
dingerkingh | Hey all - have any of you ever experienced gs maxing out the processor for a minute+ on a simple print job. Any ideas on where I could look to fix the issue? Just makes printing super slow. | 15:19.30 |
kens | Possibly your print job contains transparency, which make it not as trivial as you think | 15:20.04 |
| Without seeing your file there isn't a lto else can be said | 15:20.19 |
dingerkingh | Gotcha - this happens to any print job. Even printing from a browser at google. Would it be the driver on the cups print server that controls if gs is used? | 15:21.51 |
| Or a word document or anything for that matter. | 15:22.17 |
kens | Its up to CUPS what it uses for printing, it does mean your file will become a PDF file at some point en route | 15:22.27 |
| And printing from a browser often involves a PNG, which means you get transparency | 15:22.54 |
| But if you aren't using Ghostscript directly, we are probably not the best people to diagnose problems | 15:24.00 |
paulgardiner | pedro_: this windows build is still failing because of openssl. I can track it down myself but that seems a waste when you have it already working. Where is yours installed? Which of the versions available on the page you posted do you have? Do you have any environment vars set that might be relevant? | 15:24.10 |
dingerkingh | Ok- thanks for your responses! Last question - would virtualization have any affect? | 15:25.38 |
kens | If you mean you are running it in a VM then it'll likely be slower than running on the machine directly | 15:26.10 |
dingerkingh | Yup - thats what I meant. Thanks again | 15:27.04 |
pedro_ | paulgardiner: failing to compile/link or run? the win32ui project in the windows-alien-vs2010 specifies openssl as C:/OpenSSL/include and C:/OpenSSL/lib/VC | 15:29.49 |
| (for libeay and ssleay libs) | 15:31.14 |
paulgardiner | Thanks. Any version info - possibly in a file within OpenSSL/inc? | 15:31.47 |
| Many versions continue to be distrobuted because of API changes | 15:32.21 |
pedro_ | paulgardiner: I used: http://slproweb.com/download/Win32OpenSSL-1_0_1f.exe | 15:34.10 |
paulgardiner | Okay ta. That's what I used but allowed it to install to its default location. I guess that's the problem | 15:35.20 |
pedro_ | yup - the build files explicitly assume C:\OpenSSL - always have afaicr | 15:35.52 |
Robin_Watts | pedro: We should move that into a git submodule really. | 15:36.19 |
| The whole nightmare of "we assume that you have X and Y installed at locations Z and W" is way too hard. | 15:36.59 |
Jogux | Robin_Watts : mm, indeed | 15:38.08 |
pedro_ | nods - just the way it has been for as far as my memory can stretch. Same issues apply to directx and platform support for various builds - we haven't generally added platform-specifics into the build tree so it has been a manual setup from instructions | 15:38.33 |
paulgardiner | pedro_: I'm guessing that didn't come up as a problem for you because you were using a computer with OpenSSL already installed, but anything like that that does come up, let me know. We're looking to keep track of this sort of insider knowledge. | 15:39.53 |
pedro_ | nope, I installed it too | 15:40.09 |
| its specified in (all of the) windows projects for the various supported platforms | 15:40.57 |
| along with other dependant libs like DX9 | 15:41.13 |
paulgardiner | Where's that specified? | 15:41.23 |
pedro_ | in the project files | 15:41.30 |
Jogux | are there still issues with importing encryption binaries in/out of the USA that would prevent us adding openssl compiled libraries? | 15:41.31 |
Robin_Watts | Not for open source software. | 15:43.13 |
| oh, this is epage. ass. | 15:43.19 |
kens | was about to say.... | 15:43.29 |
paulgardiner | pedro_: Yeah I realize we can drag the information out of the build system (sometimes almost trivialy, as for this case), but still once one persons done so it's handy to keep track of that knowedge | 15:45.43 |
henrys | Jogux:go ahead we have an in-house expert for handling the government paperwork (marcosw) ;-) | 15:47.36 |
pedro_ | paulgardiner: absolutely - I think its already in the build setup stuff on the wiki; probably is worth us trying to get that live and update any dependencies. | 15:48.11 |
Jogux | henrys : :-) | 15:48.23 |
| henrys : speaking of government, we /may/ need a ERN when we submit the smartoffice update to iOS appstore | 15:49.21 |
pedro_ | Is there a preferred route for handling external dependencies? We could certainly add them to the repo and create setup scripts for new users to setup from scratch | 15:49.21 |
Jogux | henrys : (technically we probably need it for google too, but Apple may decide to actually check we have it :-) ) | 15:49.45 |
henrys | Jogux: a white tailed eagle? | 15:50.01 |
Jogux | henrys > export registration number. I think it's product specific. | 15:50.25 |
| henrys > means shovelling some paperwork around BIS's SNAP-R system | 15:51.21 |
| pedro_ : a git submodule sounds reasonable to me | 15:51.52 |
henrys | Jogux: yes this is what marcosw was doing for ghostscript and mupdf. We'll talk to him when he comes around. | 15:51.56 |
Jogux | henrys : awesome. | 15:52.03 |
| he has to finish setting up ATS first though ;-) | 15:52.25 |
| (I think he had some of the VMs booting but mace/I can't access them yet) | 15:52.43 |
paulgardiner | pedro_: not that I know of. I think the current set up is fine. I'm just making sure we have setup info written down somewhere in case bits turn out not to be in the wiki | 15:57.22 |
chrisl | How complex is the openssl build? Could the FBS not be told to build it from source when required?? | 15:57.38 |
tor8 | Robin_Watts: could we split out the mujstest generation code from mudraw please? | 15:59.44 |
paulgardiner | chrisl: certainly an option. | 15:59.53 |
Jogux | chrisl : it's a possibility. unfortunately it's not just openssl; there's a half dozen dependendies :( | 16:00.08 |
tor8 | paulgardiner: two more commits at the top of tor/master for your review | 16:00.20 |
Robin_Watts | tor8: probably. | 16:00.27 |
chrisl | Jogux: the other option would be to ensure that all missing dependencies give a sane and clear error message, and that they are expected to be in their default install locations. | 16:01.23 |
Jogux | chrisl : yep :-) | 16:01.38 |
pedro_ | chrisl - that's what we've been discussing offline | 16:01.45 |
chrisl | pedro_: then, I agree ;-) | 16:02.17 |
pedro_ | nods; that's probably one of the simpler routes, on the way to an auto-config | 16:02.43 |
chrisl | Especially with modern Windows, sticking stuff in the root directory isn't a great idea.... | 16:03.25 |
pedro_ | nods - this is just a legacy issue - its the only thing there afaik | 16:05.26 |
chrisl | Is this only for Windows, or is it for other platforms? | 16:06.13 |
Jogux | mac os x isn't nearly as bad, though still need document + graphviz + imagemagick and possibly something else. | 16:06.41 |
| s/document/doxygen/ | 16:06.49 |
| (there's mac os x setup instructions in my homedir on casper) | 16:07.26 |
chrisl | I just wondered if switching to Windows own crypto stuff would be a (future) option | 16:07.44 |
pedro_ | all of the platforms have some dependencies; its really a question of identifying them explicitly and/or installing them as part of a new build setup imho | 16:07.57 |
Jogux | chrisl : potentially, if windows essentially has built in easy to use ssl these days. (it certainly didn't in the xp days :( ) | 16:08.31 |
chrisl | Jogux: Hmm, don't know the extent/quality of what Windows comes with - not my area: I try to avoid Windows wherever possible..... | 16:09.40 |
Jogux | there seem to be a lot of people around here that say that :-) (myself included) | 16:10.01 |
kens | http://msdn.microsoft.com/en-us/library/windows/desktop/aa380255%28v=vs.85%29.aspx | 16:10.03 |
paulgardiner | tor8: possibly field.h isn't the ideal place for pdf_get_string_or_stream, although I haven't formed an oppinion of what might be better | 16:10.07 |
tor8 | paulgardiner: yeah, I wasn't sure where to put it so I put it close to where it was used | 16:10.39 |
paulgardiner | tor8: yeah, I would have probably done the same | 16:11.24 |
kens | Or better yet this URL: | 16:11.45 |
| http://msdn.microsoft.com/en-us/library/windows/desktop/bb931357%28v=vs.85%29.aspx | 16:11.45 |
| Seems to suggest SSL is built-in | 16:11.45 |
Robin_Watts | Vista and above. | 16:12.07 |
kens | caveat: I haven't used the crypto API in a *very* long time | 16:12.16 |
| Robin_Watts : but XP is no longer supported ;-) | 16:12.27 |
Robin_Watts | and, we probably don't care overly... we don't ship the windows builds to customers do we ? | 16:12.37 |
kens | hasn't a clue | 16:12.54 |
Jogux | robin_watts : who are our customers? :-) | 16:13.11 |
Robin_Watts | windows builds are just what we use to develop on. Or at least that was always the case in the past as far as I was concerned. | 16:13.20 |
| Jogux: Anyone Miles can sell to :) | 16:13.42 |
Jogux | certainly windowsrt and I believe win8 builds went to some SOT customers. but... dunno if they used the windows alien. | 16:13.42 |
Robin_Watts | Ah, well, windows 8 and rt might vaguely make sense for a touch based app. | 16:14.18 |
Jogux | unfortunately I think win8/winrt may be one of the areas bipartate did a lot of work on, so exactly what state the winrt builds in the artifex tree is like I'm not sure. | 16:16.35 |
Jogux | is trying to use win8 at the moment, and not enjoying it :-( | 16:18.10 |
paulgardiner | Jogux: let me instroduce you to vnc. :-) | 16:18.40 |
chrisl | Jogux: win8 is a whole different category that I try to avoid even more than "just" Windows..... you have my sympathy! | 16:19.08 |
Robin_Watts | Jogux: Install "Classic Shell Start Menu" | 16:19.15 |
Jogux | paulgardiner : I'm trying to figure out how well setup ats's win8 vm is. probably vnc won't help me here :-) | 16:19.15 |
| I can't even upgrade the bugger to 8.1, as it's windows enterprise and I don't have a suitable license/iso/whatever. | 16:19.40 |
| do artifex have msdn sub or anything handy like that? | 16:20.02 |
Robin_Watts | Nope. | 16:20.09 |
paulgardiner | Yeah I guessed your use of windows 8 was through need of the os and not because it was what was running on the only machine you could comfortably sit at | 16:20.25 |
Jogux | curses MS for not releasing 8.1 via windows update. | 16:21.09 |
| paulgardiner : win8 is also required for windowsrt dev work :( | 16:21.32 |
paulgardiner | tor8: so the enable/diable calls break the dependency because the app can call them rather than have the library do so during doc obj creation? | 16:22.01 |
| Jogux: Yes, that's what first forced me to use it. | 16:23.04 |
pedro_ | has to pop out for 20 minutes | 16:23.13 |
paulgardiner | To be honest, I can stand it, once I have all the desktop programs I commonly use pinned to the task bar | 16:23.38 |
Robin_Watts | the 8 -> 8.1 upgrade bit mvrhel. He updated his machines from 8 to 8.1, and that gives you a whole new product code. | 16:24.06 |
| When the machine then died, he couldn't reinstall because he hadn't written the new product code down. | 16:24.25 |
| so he ended up buying it again. | 16:24.35 |
Jogux | robin_watts : boggle! | 16:24.51 |
Robin_Watts | You have to hand MS the crown for cunning ways to make money out of 'free' upgrades. | 16:24.56 |
tor8 | paulgardiner: correct. | 16:26.37 |
paulgardiner | tor8: LGTM | 16:26.55 |
tor8 | and the drop_js happens via a function pointer so that is also a weak dependency | 16:26.55 |
| paulgardiner: fab, thanks | 16:27.00 |
| Robin_Watts: one commit at the top of tor/master for your review (split mjsgen from mudraw) | 16:27.14 |
Robin_Watts | tor8: great. | 16:30.34 |
tor8 | Robin_Watts: now I just have to fix all the MSVC projects...ugh | 16:30.55 |
ray_laptop | I have a doctor's appt (leave in about 45 min | 16:32.13 |
paulgardiner | Oh great. My now nice working windows build produces an exe that just crashes | 16:32.46 |
ray_laptop | paulgardiner: could be worse. could crash windows (like a VS installation that apparently didn't install correctly did to me) | 16:33.40 |
Robin_Watts | paulgardiner: clean rebuild? | 16:34.01 |
ray_laptop | I had to system restore (uninstall wouldn't work either) then re-instsll | 16:34.12 |
paulgardiner | ray_laptop: okay you win. That is worse | 16:34.37 |
ray_laptop | paulgardiner: if an exe crashes, it should pop up a "do you want to debug" dialog | 16:35.22 |
paulgardiner | robin_watts no. What with all the fiddling, I'll give that a try | 16:35.31 |
ray_laptop | (takes a while since it first does the worthless "checking for solution to this problem" | 16:35.55 |
paulgardiner | ray_laptop: yeah it offered to debug my vs2010-compiled exe with vs2005, but now I've restarted from the orignal solution so at least I can debug | 16:38.37 |
pedro_ | paulgardiner: did you get both patches from that branch? the second patch makes sure the app creates the temp dir it needs for the files db (which it wasn't doing before) | 16:49.21 |
| if it doesn't have a 'picseldata' directory in the system temp dir then it just drops out | 16:53.10 |
paulgardiner | pedro_: oh right. I'll pull that in. | 16:53.23 |
pedro_ | presumes all windows builds would do the same. Seemed like a very odd one, unless folks had been using installers or creating one manually | 16:53.57 |
paulgardiner | Don't assume I have any memory of how this works. :-) | 16:54.09 |
Jogux | pedro_ : mmm, very strange. | 16:54.23 |
kens | memory ? I've heard of that.... | 16:54.24 |
pedro_ | paulgardiner: basically someone just using File_open rw rather than doing File_createPath first | 16:55.50 |
| kens: I know what you mean ;) | 16:57.09 |
paulgardiner | pedro_: so maybe I should be building app/simple-install, rather than app/simple, at least initially to create the installer that will create picseldata?? | 16:57.24 |
pedro_ | isn't sure if there actually is an installer for it | 16:57.52 |
paulgardiner | Second ? was accidental not for emphasis | 16:58.05 |
pedro_ | :) | 16:58.23 |
Jogux | it'd be interesting to see what ats is building perhaps (although I think what ATS is building is just some random choices, rather than what ATS actually was building on the real live SOT system :-( ) | 16:58.50 |
paulgardiner | pedro_: so is everything the exe needs created by the exe, provided the temp dir exists? | 16:58.58 |
pedro_ | (I guess the win8/RT builds will do) | 16:59.01 |
Jogux | paulgardiner : /which/ exe ;-) | 16:59.20 |
| if it's ue2fileviewer you'll need some mcfs too | 16:59.27 |
pedro_ | paulgardiner: yes, all resources are bundled, but it creates a db to remember files you've opened etc | 16:59.32 |
| jogu> they're there with the exe | 16:59.50 |
Jogux | may have misread paul's question | 17:00.11 |
pedro_ | may have too ;) | 17:00.34 |
paulgardiner | I may have miswrote it too | 17:00.50 |
pedro_ | we have all bases covered then ;) | 17:01.01 |
Jogux | :-) | 17:01.07 |
paulgardiner | I even miswrote miswrtten | 17:01.39 |
Robin_Watts | I think she used to be my primary school english teacher. | 17:03.35 |
paulgardiner | :-) | 17:04.16 |
pedro_ | if we want to sort out build dependency warninga and build an installer I'm happy to stick that on my list - was just getting this build going to help me debug some of the layout issues initially | 17:04.50 |
paulgardiner | Nah, I think it'll be fine how it is, provided we document it (or already have it so - as we suspect) | 17:06.31 |
Jogux | so. afaict all the windows ats images are unable to actually do builds. the win8/vs2012 one falls over at the beginning of the alien build; can't find picselwin32.lib. | 17:06.43 |
pedro_ | cool | 17:07.08 |
Jogux | unless I've missed some parameter or something. | 17:07.26 |
pedro_ | did you do --windows8? | 17:07.47 |
Jogux | no... do you think I should or is that a metro build? | 17:08.09 |
| meh, I'll try anyway :-) | 17:08.15 |
pedro_ | it may only be metro and winphone8 that build (as opposed to vanilla ue2dispman) bicbw | 17:08.41 |
Jogux | that is certain possible :( | 17:08.54 |
pedro_ | there's a bunch of code to bundle specifically for w8 builds in win32_postlink.py | 17:09.07 |
mvrhel_laptop | robin_watts I am going to have a commit a bit later for you to review to fix the mupdf winrt stuff. the file streaming is broken | 17:09.16 |
pedro_ | --windows8 sets up the relevant defines afaics | 17:09.20 |
Robin_Watts | mvrhel_laptop: Yeah, sorry, I forgot that the change would affect you. Sorry. | 17:09.44 |
Jogux | pedro : neat. presumably that won't get me a win32ui.exe with all the nice debugging features though (which is what I was really after :-S ) | 17:09.44 |
Robin_Watts | mvrhel_laptop: Does the API change make sense? | 17:10.34 |
| essentially there is no stm->buffer any more. | 17:10.50 |
| and 'read' is replaced by 'next', which is expected to fill 'some' data into a buffer, point stm->rp at the first byte of the buffer, stm->wp at the byte after the end, and then exit with return *stm->rp++; | 17:11.47 |
mvrhel_laptop | robin_watts I will let you know when I wade through it. thanks for the explanation about the changes | 17:12.25 |
kens | OK goodnight all | 17:14.18 |
pedro_ | Jogux: it should be fairly straightforward for me to get the 2012/win8 stuff building ue2dispman from what I've seen | 17:20.45 |
Jogux | pedro_ : nod. | 17:21.13 |
Jogux | is just going to setup win7 + vs2005 for now I think; that's likely very close to what SOT were using so should in theory work "best" :-) | 17:21.52 |
| this is a completely clean VM too, so it'll give me a chance to fully discover & document all the dependencies | 17:23.33 |
Robin_Watts | Having just been to Texas: http://www.youtube.com/watch?v=ze1TTNillls | 17:52.34 |
| I'm looking at setting up twiki.ghostscript.com now. | 19:14.29 |
| I vaguely have the twiki up and working. I've nobbled it so it needs a password so it's not public currently. Will fiddle some more later/tomorrow. | 20:30.49 |
| ray_laptop: For the logs: We need to make twiki.ghostscript.com point to the same address as casper. | 20:38.41 |
mvrhel_laptop | robin_watts: you still up? | 20:43.53 |
| nevermind I think I have this figured out now | 20:47.32 |
| :( the stream stuff in winrt has some issues | 22:22.17 |
| ok got it working now | 22:34.45 |
Jogux | Robin_Watts : oo, awesome, thanks. | 23:26.40 |
mvrhel_laptop | robin_watts: for the logs. the fixes for the winRT stuff are on my repos | 23:51.58 |
| if you can review that would be great. | 23:52.07 |
| Forward 1 day (to 2014/03/21)>>> | |