| <<<Back 1 day (to 2015/07/27) | 20150728 |
Robin | Morning kens | 08:13.09 |
kens | Mornign Robin, how's your holiday going ? | 08:13.25 |
Robin | tor8: (For the logs) I'm off to bed now. Will be back in about 8 hours if you have questions about anything I've done in your absence. Mainly just gproof though. | 08:13.54 |
| kens: Going well thanks. | 08:14.07 |
| kens: Did you get tickets for Chicago? | 08:14.24 |
kens | Chore for today :-( | 08:14.32 |
| I gather Helen and STella are talking about December meeting, no doubt I'll be informed (eventually....) | 08:14.55 |
kens | is steadily working through the emila nad IRC backlog | 08:15.25 |
Robin | heh. H has been in touch with most of the WAGs I think :) | 08:15.26 |
kens | Fortunately I saw Miles's mail while we were away, I'd taken the laptop along to dump photos onto | 08:15.51 |
| Robin were you able to book extra legroom seats on Virgin and still choose you seat allocation ? | 08:18.04 |
Robin | Yes. | 08:18.57 |
| Choosing seats costs 25quid a flight. | 08:19.12 |
kens | Ah, that wasn't possible before, good news. Yeah I knew about the allocatoin cost :-( | 08:19.23 |
Robin | Choosing extra legroom seats costs 40 quid a leg. | 08:19.26 |
kens | Yes I saw your email | 08:19.34 |
| Or IRC I forget which, I'm losing track a bit here | 08:19.46 |
Robin | yeah, given we have to pay 50, paying 80 seems less unreasonable. | 08:19.48 |
| IRC, I suspect. | 08:19.51 |
| bedtime for me... | 08:19.59 |
kens | Have a good night | 08:20.04 |
Robin | ta. | 08:20.08 |
kens | hopes to be staright by your breakfast time :-) | 08:20.25 |
| chrisl ping | 08:23.17 |
| ah forget that | 08:23.51 |
| What is it with the crazy Virgin website ? It won't let me book a flight, keeps poionting to my date of birth and saying 'check the highlighted areas to see what went wrong'. Oh and if I change my DOB to yhe 8th its happy. I wonder if they've stuffed up my details...... | 08:37.37 |
| I wonder how RObin managed to book extra legroom, I can't see a way to do it in the website. | 08:38.08 |
chrisl | kens: I think you have to do the extra legroom when you select your seat | 08:40.58 |
kens | Oh I hadn't got that far, thanks | 08:41.08 |
| Still truggling with my date of birth. Looks like I have to set it up wrong, then go back and change it, then its happy O.O | 08:41.40 |
chrisl | I had not dissilimar problems when I booked for the last staff meeting | 08:42.12 |
kens | Mad, totally mad..... | 08:42.27 |
| Hmm, it wants me to pay before offering a seating option, oh well. Not exactly a disaster if I end up with regular seats I guess | 08:44.38 |
| Oh, nad going via the old stire doesn't complain about my DOB..... | 08:44.54 |
chrisl | It demands payment when you go to select your seat | 08:47.05 |
kens | OK seems reasonable, I jst closed the browser by accident, so I get to go back and start again..... I must be tired | 08:47.29 |
chrisl | Or possibly just a comment on the VA website...... | 08:48.01 |
kens | And when I try to choose seats, the site crashes, excellent.... | 08:58.15 |
| Hmm, there are some exit row seats available..... | 09:00.29 |
| OK I@mfinally done. | 09:04.22 |
| chrisl on the plus side, the directory changes seem to have worked with no problems for me. | 09:11.24 |
| Can I ask what the 'old-stuff' directory is for ? | 09:12.43 |
chrisl | kens: that's just a place holder for some stuff I need to double check, update or remove before the release | 09:15.42 |
kens | Ah OK | 09:15.48 |
chrisl | Rarely used things like the Metro build option | 09:16.07 |
kens | Should .gitignore be included in Git ? | 09:16.11 |
chrisl | I think so yes, so we all start with the same contents | 09:17.00 |
kens | OK | 09:17.05 |
chrisl | I *think* .gitignored is, itself ignored, so when you clone a repo you get the default for the repo, then you can customise it as you prefer | 09:18.14 |
kens | If I modify .gitignore then git gui whinges that I've changed it | 09:18.38 |
chrisl | Oh, well, maybe we need to put .gitignore in .gitignore..... | 09:19.05 |
kens | :-) Its not a big deal | 09:19.13 |
tor8 | morning | 09:23.16 |
| kens, chrisl: looking at flights to bahamas, my best option seems to be via the BA flight from London to Nassau | 09:23.48 |
kens | Could be, I'm inclined to fly BA from here because we can book right through, even though we have to change in Miami and fly a US airline | 09:24.26 |
chrisl | I hadn't even looked..... | 09:24.57 |
tor8 | changing in a non-US destination seems like a better idea to me :) | 09:25.40 |
kens | tor8 yes agreed, but I couldn't see such a route. Admimttedly I was in Ireland at the time and the web sites all wanted to offer me flights from Dublin..... | 09:26.29 |
tor8 | kens: the LHR-Miami flight ends up at 35 hours total for me | 09:27.00 |
| whereas the Nassau is only 17h | 09:27.06 |
kens | Oh wow.... | 09:27.08 |
| Both are pretty bad | 09:27.15 |
tor8 | it's an overnight miami stay | 09:27.29 |
kens | Oh | 09:27.40 |
| Where is Nassau anyway ? | 09:27.46 |
tor8 | 45 minutes away from freeport | 09:27.54 |
| by plane | 09:27.59 |
kens | OK that's the same as Miami | 09:28.07 |
tor8 | it's the big city in Bahamas, IIRC | 09:28.15 |
kens | I'd prefer not to do immigration at Miami | 09:28.17 |
| SO that's Lynden Pinding International Airport I guess | 09:28.59 |
| Ah right, its the biggest (ie only real) international airport | 09:29.24 |
tor8 | kens: yeah. not many flights, but I think avoiding a change of planes in miami is worth quite a bit | 09:30.44 |
kens | I agree completely | 09:30.52 |
pavan | Hi kens | 09:31.04 |
kens | I don't know if we'd have to do 2 sets of immigration, but I;d guess that would be a yes | 09:31.15 |
| Hi | 09:31.17 |
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. | 09:31.17 |
tor8 | if going through Miami, I would expect so, yes | 09:31.39 |
pavan | fz_open_document(ctx, CString); is throwing access violation exception ? | 09:31.48 |
kens | WHich would be a real pain | 09:31.57 |
| pavan if that's a question addressed to me, I have no clue, I am not a MuPDF developer | 09:32.15 |
pavan | am i in correct place ? | 09:33.04 |
kens | Yes | 09:33.11 |
| But you'll have to gove more informatoin I expect. | 09:33.25 |
| So BA, route LHR->MIA->FPO is £1500 | 09:34.09 |
tor8 | pavan: are you using the C library, and if so what is a "CString"? | 09:34.25 |
| kens: so for chicago, we're all staying until sunday I take it? | 09:34.46 |
pavan | that is file path... | 09:34.50 |
kens | tor8 yes, we need the Saturday night | 09:34.56 |
| CString looks like a C++ object to me | 09:35.12 |
tor8 | kens: yeah, my flights are not fun. it's a 22.30 flight out on sunday | 09:35.24 |
kens | THats | 09:35.33 |
| Sorry, that's worse than ours | 09:35.39 |
pavan | i have converted platform string into C string | 09:35.41 |
tor8 | I'm going to ask Miles for an extra night so I can use the hotel room during the day | 09:35.41 |
kens | I thnk that's sensible | 09:35.53 |
tor8 | the chicago airport terminal 5 is way too small to loiter all those hours :( | 09:36.24 |
kens | I can't find a BA flight directo to Nassau, they all seem to go to Miami, with the single exception of one via JFKL | 09:36.37 |
pavan | Platform::String^ Path = "filepath"; const wchar_t *W = Path->Data(); int Size = wcslen(W); char *CString = new char[Size + 1]; CString[Size] = 0; for (int y = 0; y<Size; y++) { CString[y] = (char)W[y]; } | 09:36.47 |
kens | So that's C++, not C | 09:37.00 |
pavan | yes | 09:37.39 |
| now why it is throwing exception ? | 09:38.10 |
kens | I have no idea | 09:38.22 |
| MNost likely you are passing an invalid value for ctx I should thnk | 09:38.38 |
tor8 | you'd need to get a stack trace from gdb to help | 09:38.38 |
pavan | Actually i got correct pointer for context.. | 09:39.41 |
| i just used like this fz_new_context(NULL, NULL, FZ_STORE_DEFAULT); | 09:39.54 |
kens | Then I suggest you follow the call with a debugger and see where its throwing the seg fault | 09:40.13 |
pavan | That is the problem actually i created dll for that... | 09:40.48 |
kens | And have you actually looked at the value of ctx, or are you just assuming its correct ? | 09:40.56 |
pavan | i can't debug on dll.... | 09:40.56 |
| if it is not NULL probably it is a valid one...isn't it ? | 09:41.35 |
kens | Well I certainly can't debug it over IRC | 09:41.35 |
| No.It could be random | 09:41.47 |
| uninitialised memory is not required to be 0 | 09:41.57 |
pavan | i have initialised with ) | 09:42.44 |
| 0 | 09:42.49 |
kens | Its no use you asking me what's wriong, and when I make suggestions you telling me 'no, I already thought of that'. I cold spend all day playing that game. | 09:43.59 |
| THs is what I meant when I said you would have to supply more information | 09:44.19 |
| tor8 skyscanner says there are no direct flights from the UK to the Bahamas | 09:46.03 |
tor8 | kens: I couldn't find any today on the nassau airport web site | 09:47.30 |
| I expect they're only seasonal | 09:47.34 |
kens | Yeah I did wonder if that might be it | 09:47.44 |
tor8 | it's flight BA 253 | 09:48.03 |
kens | If we have to go via the US anyway, then Miami and straight to Freeport seems the sensible option | 09:48.04 |
tor8 | kens: yes. depending on the actual date I can get 18h flights via miami, but on the 4th that I just picked at random at first, all the miami flights are 30h-ish | 09:48.47 |
chrisl | The BA flight to Nassau only seems to be on certain days - I had shift the dates around to find it | 09:49.10 |
kens | I was looking at going out on 30/11 | 09:49.22 |
| I'd be happy to fly to Nassau if the BA site would tell me when it goes :-) | 09:49.42 |
| Looks like myabe wednesday and Friday | 09:49.57 |
chrisl | Yeh, I think it came up when I selected Dec 4th | 09:50.29 |
kens | It departs on the 2nd at least | 09:50.42 |
tor8 | dec 1st has one Nassau flight | 09:50.46 |
kens | Prices look fair | 09:51.21 |
| Direct flight back on the 6th | 09:51.45 |
tor8 | my other option that day seems to be airberlin, and routehappy.com tells me they only have 30" pitch seats :( | 09:51.54 |
kens | total £730 | 09:51.58 |
| I was planning to leave this to Stella and Helen :-D | 09:52.23 |
tor8 | kens: good choice :) My SO is not interested in flying that far, so I was hoping to pester sebras until he caves and keeps me company :) | 09:53.26 |
kens | Flying on Tuesday and back on Sunday looks good if we go direct to Nassau. Of course we have to change for a flight to Freeport, but I bet they are frequent | 09:53.30 |
| tor8 that would be fun | 09:53.41 |
| Since I missed sebras last time he was in the UK | 09:53.52 |
sebras | kens: exactly! :) | 09:55.19 |
kens | Hmm, no direct flight back on Monday. Whch means either staying 5 nights or 7 nights, instead of 6. | 09:55.38 |
sebras | tor8: what are the dates? | 09:56.05 |
kens | sebras, at the moment flexible around the 5th and 6th XDecember | 09:56.24 |
| sorry I mean 4th and 5th | 09:56.50 |
sebras | kens: tor8: my calendar for december seems really empty (no surprise). | 09:57.18 |
kens | Fancy some winter sunshine ? | 09:57.30 |
| Top up the vitamin D for us Northern types.... | 09:57.43 |
tor8 | sebras: I expect that comes down to which field trips Helen's planned that we can crash :) | 09:57.45 |
sebras | kens: why not. can you start by sending me som _summer_ sunshine first..? | 09:57.58 |
kens | I could use some of that too | 09:58.07 |
tor8 | sebras: that's not true. we had at least one day of sunshine! | 09:58.17 |
sebras | turns head | 09:58.25 |
kens | Ws on holiday in Ireland last week and we had steretypically Irish weather; rained every day | 09:58.28 |
sebras | oh.. the sun _is_ shining. I just need to look out the window and not at the computer screen. | 09:58.44 |
| but it _is_ a really cold summer here though. | 09:58.55 |
tor8 | sebras: yeah. it's not at all feeling like end of july today. more like october. | 09:59.21 |
sebras | kens: I'm going to dublin for a conference in october. I expect rain... | 09:59.31 |
kens | And lots of it :-) | 09:59.41 |
| I was amused by the Irish Met Office trying to persuade people that its not that bad "Tested on an hourly basis, 2 out of 3 hours are dry". Which I read as 'it rains 8 hours out of 24.....' | 10:00.46 |
| Looks like flights from Nassau to Freeport run at £132 for a round trip | 10:01.38 |
| Flying with BahamasAir | 10:01.49 |
tor8 | kens: you could try booking through expedia, though from experience their rates are usually the same as if booking with the airlines directly | 10:05.57 |
kens | Sure, I was just wondering what the frequency and cost would be like. | 10:06.25 |
| Looks like we could get a flight out of Nassau 2 hours after landing. | 10:06.38 |
| I suspect we'll have to wait for Robin and Helen to return before we can book anything to be honest. | 10:07.49 |
tor8 | kens: Yeah. I just wanted to point out that there was an option you might otherwise have overlooked. | 10:08.34 |
| given how obscure it is to find :) | 10:08.45 |
kens | tor8 yes I hadn't found that flight | 10:08.50 |
| TO be fair, I hadn't had a chance to look since coming home yet | 10:09.07 |
| It looks like a good bet to me, 2 sets of immigration would be a pain, and Miami immigration is always bad | 10:09.33 |
| It may even work out cheaper | 10:09.49 |
| In terms of excursions I looked at beach rides and there only seems to be 1 place, and they ride Western (no surprise I guess), which puts me off. I was wondering aout learning kite-boardng while we're there, that looks like fun | 10:10.52 |
sebras | tor8: I have discovered a design flaw in how we do repair I think. | 12:11.54 |
| tor8: consider the case where pdf_load_object() is asked to load object 8 0 R | 12:12.15 |
| tor8: now, in the file 8 0 obj has been replaced by 0 0 obj, leaving the rest of the object dict intact. | 12:12.42 |
| tor8: so we call pdf_cache_object() | 12:12.52 |
| tor8: which will call pdf_parse_ind_obj() which will now load something called 0 0 R, and populate rnum and rgen both with zeroes. | 12:13.29 |
| tor8: because we are not repairing and rnum (0) and num (8) are not equal we will drop the loaded object and enabled repair | 12:14.10 |
tor8 | sebras: "and enabled repair"? | 12:14.53 |
sebras | enable repair | 12:15.00 |
| tor8: there's a flag called try_repair. | 12:15.14 |
| in pdf_cache_object() | 12:15.24 |
tor8 | ah, that's robin's doing so I'm not up to date on the details | 12:15.29 |
sebras | right. | 12:15.37 |
| ok, anyway, so we pdf_repair_xref() will not fail because it will simply ignore object 0 0 obj in the file. | 12:16.00 |
| tor8: because its object number if out of range | 12:16.13 |
| tor8: then pdf_cache_object() will loop back and attempt to call pdf_parse_ind_obj() again, which again will load object 0 0 R | 12:16.49 |
| (since there is not anything wrong with the file format per se) | 12:17.02 |
| tor8: and for the same reason we will again drop the loaded object. | 12:17.43 |
| tor8: but this time around we have attempted a repair once before so we will not try it again. | 12:17.59 |
tor8 | sebras: hm, so the repair_xref call doesn't initialise the '8' object slot to free? | 12:18.06 |
sebras | tor8: no, it is left as 'n' | 12:18.18 |
tor8 | well, that's not right. | 12:18.25 |
sebras | tor8: no, but the problem appears at a later stage. | 12:18.34 |
| tor8: next we call pdf_set_obj_parent() which will succeed (even though x->obj == NULL) | 12:19.00 |
tor8 | I would have expected pdf_repair_xref to zap the existing entries when scanning the file | 12:19.04 |
sebras | tor8: the reason being that NULL < PDF_OBJ__LIMIT and for those there is no parent to be set. | 12:19.44 |
| tor8: but still there is no problem. | 12:19.55 |
| tor8: so pdf_cache_object() returns. | 12:20.05 |
| tor8: returning back to pdf_load_object() which asserts that entry->obj != NULL. | 12:20.25 |
| tor8: and _then_ we fail. | 12:20.30 |
| tor8: I'm not sure that setting the '8' object slot to 'f' would have helped...? | 12:20.56 |
| tor8: or maybe the second time around x->type == 'f' and then we'd end up calling pdf_new_null() instead of pdf_parse_ind_obj() a second time. | 12:21.51 |
tor8 | sebras: if we set it to 'f' (as we should because we can't find any 8 0 obj in the file) | 12:22.02 |
| then after the object_updated: try-again-later label, we'd set x->obj = fz_new_null() | 12:22.25 |
sebras | tor8: yeah, I just realized that. | 12:22.35 |
tor8 | I think the problem is that we're keeping known entries when repairing, but that won't work when the old entries are bust | 12:23.06 |
| but I think it's best to ask Robin_Watts about this, he said he'd check the logs when he wakes up this evening (he's on holiday) | 12:24.23 |
sebras | tor8: I seem to think athat the problem is at around line 475 in pdf_repair_obj() | 12:24.33 |
| no, we do only iterate over the list of actually found objects (i.e. listlen will not be incremented for the ignored object. | 12:25.51 |
| tor8: got is. | 12:27.21 |
| tor8: got it. | 12:27.27 |
| tor8: in pdf | 12:27.35 |
| in pdf_cache_object() when we discover that we got object 0 0 R instead of the exepcted 8 0 r we drop x->obj for object 8 0 R. | 12:28.16 |
| when we do that we should simultaneously mark that slot as being 'f' and reset all other parts of that slot. | 12:28.48 |
| _then_ we repair, and if we do find the object: great! | 12:29.05 |
| if we don't, we'll just end up calling pdf_new_null() at the next iteration. | 12:29.21 |
tor8 | sebras: yeah, so any xref entries not found (i.e. not in the list/listlen in pdf_repair_xref) are untouched, where they really ought to be zeroed | 12:35.40 |
sebras | tor8: yes. | 12:36.24 |
tor8 | I would have expected to see a zeroing pre-pass in pdf_repair_xref, that takes care to drop the already loaded objects | 12:36.45 |
sebras | tor8: I guess it depends on whether you still trust your original xref after having discovered a problem with it. | 12:36.55 |
tor8 | but maybe Robin want to preserve already loaded objects | 12:37.14 |
sebras | tor8: what would happen if pdf_repair_xref() would actually free all objects and they are referenced elsewhere? | 12:37.48 |
tor8 | we don't actually touch the entry->obj entries | 12:37.50 |
sebras | would something break? | 12:37.52 |
| tor8: not unless we end up in pdf_repair_obj_stm(), no. | 12:38.19 |
tor8 | sebras: ah, it could break badly | 12:38.26 |
kens | We had a PDF file a while back that had duplicate objects with the same object and generation number. And teh *first* one was hte correct one. Previously we've found the same where the *last* one was correct. Ths file only worked if you trusted the xref in the areas wehere it was not incorrect. | 12:38.32 |
tor8 | since we might end up calling pdf_repair_xref from inside a pdf_resolve_indirect, so at any time really | 12:38.47 |
| which means anyone holding a transient (inside their own function) reference to a pdf_obj could end up holding a freed object | 12:39.32 |
| still, I'd expect zeroing the entry->type to 'f' would not hurt | 12:39.58 |
sebras | tor8: right. that's a bad idea then AIUI. | 12:40.03 |
| tor8: the patch I whipped up at least sets the known missing object to 'f'. | 12:40.27 |
| tor8: the remaining objects will get the same treatment once we try to pdf_load_object() them I guess. | 12:40.52 |
tor8 | I think your approach of only zapping it at the point of caching could catch all bad cases | 12:41.05 |
sebras | tor8: and we'd probably handle kens' case pretty well too. | 12:41.38 |
tor8 | yeah, since you set it when the error is detected, whether the repair has been done already or not | 12:42.14 |
sebras | tor8: ok, so another patch over at sebras/master | 12:42.33 |
| tor8: /me added a period at the end of the commit message | 12:42.57 |
tor8 | I'm not sure it'll cope with kens case | 12:43.03 |
kens | That was a pretty horrible case.... | 12:43.16 |
tor8 | sebras: :) | 12:43.17 |
sebras | yeah, I'm also still thinking about it. | 12:43.20 |
| tor8: the rule is to NOT have them at work. sorry, I've been tainted. | 12:43.37 |
kens | I'll try and remember the bug number and you can check it out, my memory is hazy | 12:43.40 |
sebras | tor8: actually I think mupdf would alway opt for the _last_ object with the same object number. | 12:44.44 |
kens | GS did too, that's why it was wrong | 12:44.56 |
sebras | tor8: I don't think we'd even get a warning | 12:45.05 |
tor8 | ahh! mouse focus and ^W ... I've been using my windows machine for too long :) | 12:47.18 |
sebras | tor8: ^W is close window... but what is it in windows? delete word? | 12:48.08 |
tor8 | sebras: click-to-focus... | 12:48.46 |
sebras | oh. | 12:48.55 |
kens | tor8 sebras, the file with the problem I mentioned is attached to bug #696084, the GS bug fix commit is : | 14:01.03 |
| http://git.ghostscript.com/?p=ghostpdl.git;a=commit;h=326a116687ee35db6d1bdcc54f2a8b2fb4458878 | 14:01.03 |
| which contains my ramblings about the problem. | 14:01.03 |
| That particular file may not cause MuPDF a problem, the xref entry that has a problem is not actually used in the file, so its never consulted. The reason its a problem for GS is that we sanity check the xref before we start, because we can't easily fix it afterwards. | 14:03.56 |
henrys | 5 minutes to the meeting | 14:24.39 |
jogux | kens: tor8: (or anyone else) - just wondering if you are able to make a quick informed 'guess' why http://bugs.ghostscript.com/show_bug.cgi?id=696118 might be wrong in ghostdocs (there's a screenshot attached; you'll need to be logged into bugzilla to see the bug). mupdf gets it right... | 14:25.11 |
henrys | has SO marketing on the brain, that's been my life for the last week and I can't shake it. | 14:25.14 |
kens | jogux will need a few minutes, will check later | 14:25.29 |
jogux | henrys: the whole week? gees :( | 14:25.49 |
| kens: many thanks :) | 14:25.52 |
henrys | jogux: just wait for your meeting ;-) | 14:26.22 |
chrisl | jogux: transparency would be my guess...... | 14:26.27 |
jogux | hehe, sorry, skipping the gun :) | 14:26.32 |
| I thought mupdf didn't do transparency? | 14:26.38 |
chrisl | It does most of PDF transparency - only a few edge cases it misses | 14:26.58 |
jogux | ah, right. | 14:27.11 |
chrisl | I guess the "shading" around the text are soft masks | 14:27.36 |
henrys | kens, chrisl - we are short robin so... | 14:28.27 |
tor8 | jogux: without looking at the PDF contents, I'd hazard a guess at softmasks or fancy transparency blending to get the white fuzzy blob behind the text | 14:28.39 |
sebras | agh! is "windows technical support" calling you guys in .uk and .us as well? three times today. :-( | 14:28.39 |
tor8 | sebras: beware the scammers | 14:28.59 |
sebras | tor8: I know, unfortunately my caller id display broke so I can't see who is calling. | 14:29.53 |
henrys | in the US they legally are supposed to respond to "Take me off this list" but they don't | 14:30.24 |
tor8 | you should get what robin said he got, a whitelisting kind of service | 14:30.26 |
chrisl | I got that too - it's great! | 14:30.41 |
sebras | kens: bugs 696084 does not have any file attached though... | 14:30.57 |
| kens: is it private? | 14:31.00 |
kens | Oh rats | 14:31.04 |
tor8 | henrys: what these jokers do is illegal to begin with, I don't think they care too much about no-call lists | 14:31.05 |
kens | I'll try and find the file | 14:31.08 |
marcosw | morning | 14:31.41 |
sebras | kens: it would be nice to verify that mupdf parses it correctly. | 14:31.42 |
jogux | I reckon we don't do softmasks in ghostdocs so I sounds like a good candidate. | 14:31.43 |
henrys | item 1 for the meeting is for kens : we do need a good explanation for Aaron why we can't generate /WinAnsi - "hell no" is a little terse. Can you write something up? | 14:31.58 |
kens | sebras I suspect it will be OK with MuPDF because the breoken xref is never actually used | 14:31.59 |
| henrys there are 2 open bug reports | 14:32.11 |
| with extensive write ups | 14:32.17 |
| We cna probably make *this* file work, but not replicate what Acrobat does | 14:32.40 |
| And making this file work will be a substantial amount of effort | 14:33.06 |
henrys | kens: would it be best for you to simply enter negotiations with aaron? | 14:33.08 |
kens | Probably not, since my inclination is to say 'no' loudly | 14:33.25 |
| How much effort (ie how much of my time) are we prpared to devote to ths ? | 14:34.00 |
chrisl | jogux: Ghostcript's output with transparency disabled: http://ghostscript.com/~chrisl/stuff003.png | 14:34.19 |
henrys | kens: none I just want a good explanation to send him that is polite and reasonable. | 14:34.37 |
jogux | chrisl: brilliant... thanks :) | 14:34.39 |
kens | the simple explanation is 'we don't do re-encoding to winansi' | 14:34.52 |
chrisl | jogux: NP, I like proving myself right ;-) | 14:35.04 |
kens | While we could do so, it is not a requirement of the spec, and would involve us in signifacnt effort, and would likely lead to a lengthy bug tail | 14:35.27 |
henrys | kens: that's all I need thanks. marcosw and I did not know that. | 14:36.10 |
kens | NB we cannot replicate Acrobat's behaviour re-encoding to WinAnsi with certain types of input. | 14:36.36 |
henrys | welcome back tor8 we've saved a list of projects for you while you were gone ;-) | 14:37.58 |
tor8 | henrys: uh oh... | 14:38.10 |
henrys | tor8: there is really only one big project right now and that is the dreaded XFA. | 14:38.37 |
tor8 | henrys: then I stand by my point... | 14:38.58 |
henrys | tor8: I have jeong's implementation in C++ in my home directory, XFA seems to be isolated in a folder, can you at least have a look and see if that can be used as a guide ... | 14:40.15 |
rayjj | henrys: that's looking over the code from jeong as the first step, right ? | 14:40.18 |
henrys | rayjj: yeah have you already looked at it? | 14:40.41 |
rayjj | C++ and tor don't mix ;-) | 14:40.44 |
| henrys: nope. | 14:40.50 |
| If it's "real" C++ I stay as far away as possible as well :-) | 14:41.27 |
tor8 | henrys: I could take a look. I also need to study our current forms architecture, it's been a while. | 14:41.39 |
| henrys: where in your home directory? | 14:41.54 |
henrys | rayjj: we're using it as a guide not a library | 14:42.05 |
rayjj | are we thinking that we will re-do it in C ??? | 14:42.35 |
henrys | tor8: pdfv-[hash] | 14:42.45 |
mvrhel_laptop | that seems a bit extreme | 14:42.55 |
| but then again tor8 did write the mujs in a few days... | 14:43.27 |
henrys | tor8 will not allow C++ in mupdf I wasn't even going there. | 14:43.38 |
paulgardiner | We've had C++ in mupdf before, but carefully hidden behind a C interface. Did that to use v8. | 14:44.48 |
chrisl | I'd be very surprised if the XFA stuff was usable as a library, in the way v8 was | 14:45.34 |
tor8 | paulgardiner: and now you know why I wrote mujs ;) | 14:45.36 |
paulgardiner | I was just thinking that. :-) | 14:45.49 |
henrys | next item is about UI's for mupdf and gsview. We have hired a designer to do work on SO and the thought is being batted around that this person could have a look at mobile mupdf and gsview. | 14:46.16 |
| (duck) | 14:46.43 |
rayjj | henrys: shouldn't we wait and see how well the SO project turns out ? | 14:47.09 |
mvrhel_laptop | henrys: I suppose it can't hurt | 14:47.10 |
| to get the feedback | 14:47.26 |
henrys | rayjj: that's a reasonable point | 14:48.12 |
| did you want to weigh in on that jogux ? | 14:48.40 |
tor8 | henrys: there's an awful lot of code in that project, but I take it the whole does a lot more than just XFA | 14:49.10 |
henrys | tor8: you probably just need to look in the xfa directory. that's his whole epapyrus product | 14:49.45 |
jogux | henrys: I think the designer would be up for that (as it happens I already showed him the current mupdf ui when explaining about artifex etc.), and it would definitely help. I can't make the cost / benefit call of course. | 14:50.12 |
henrys | rayjj point is well taken but I don't know that "management" wants to wait. They seem to be in a hurry with all this stuff. | 14:51.13 |
jogux | I would imagine they wouldn't have bandwidth to look at mupdf/gsview before the SO project is to the point we can see results, so rayjj's point is reasonable. | 14:51.17 |
henrys | jogux: okay then that solves it. | 14:51.34 |
tor8 | henrys: heh. when did Jeong port mupdf to C++? | 14:53.07 |
henrys | anything else for the meeting, I'm good. | 14:53.18 |
| ? | 14:53.23 |
marcosw | There are a couple of open customer bugs I'm not sure who to assign to: | 14:53.51 |
| http://bugs.ghostscript.com/show_bug.cgi?id=696107 | 14:54.02 |
henrys | tor8: I don't know he did, I assumed he just used mudpf and built up C++ additions around it. | 14:54.08 |
marcosw | http://bugs.ghostscript.com/show_bug.cgi?id=696028 | 14:54.15 |
| anyone want to volunteer? | 14:54.28 |
tor8 | henrys: looking through the XFA I saw he had a pdf interpreter in C++, and looking through that I see a shared genealogy with mupdf, but in C++ | 14:54.44 |
| same code structure and naming, but heavily rewritten in C++ | 14:55.03 |
henrys | tor8: why would he do that? makes upgrades terribly difficult | 14:55.27 |
chrisl | marcosw: isn't question on 696107 about whether we should throw an error or not? | 14:55.32 |
mvrhel_laptop | henrys: So I should have my color management changes for gproof in gs wrapped up today or tomorrow at the latest. Then I need to do a bit of testing with it using mupdf to make sure it all looks reasonable. | 14:55.37 |
rayjj | marcosw: bug 696028 is intermittent. I don't think there is anything we can do | 14:55.45 |
tor8 | henrys: a morbid fascination with C++? | 14:56.17 |
| it makes upgrades impossible | 14:56.21 |
rayjj | marcosw: unless you can reproduce it. BTW, thanks for closing the other windows bug | 14:56.30 |
tor8 | there's not a single line of code that's the same, but all the lines of code are originally mupdf C | 14:56.44 |
chrisl | Goes wrong 1 in 100 times - great :-( | 14:56.47 |
jogux | tor8: oh, did you see there's an 'Artfiex' mupdf auto-mirror on github now that ChisL setup - https://github.com/ArtifexSoftware/mupdf - we were talking about mirroring the submodules too, but that would mean a) changing the git submodule urls to be relative (seems like a reasonable thing to do anyway?) and b) renaming them as github doesn't let us has directories (ie. thirdparty/)... [partly this came about because I created a cocoapod for mupdf | 14:57.14 |
| was worried about the load/bandwidth requirements that might put onto casper] | 14:57.15 |
rayjj | kens: why did you have to notice that this was a customer bug (696028) ??? | 14:57.31 |
chrisl | marcosw: you can assign http://bugs.ghostscript.com/show_bug.cgi?id=696107 to me | 14:57.52 |
tor8 | jogux: right. the thirdparty submodule urls are a problem. | 14:57.55 |
| jogux: can submodule url be relative? I don't think we want to change them now, because then checkouts to old versions will fail. | 14:58.33 |
jogux | tor8: http://git.ghostscript.com/?p=user/joseph/mupdf.git;a=commit;h=f94735aa9477c384c507aaca02f665e0854d3e76 | 14:58.36 |
marcosw | chrisl: thx, done | 14:58.39 |
tor8 | jogux: right. if that works, perfect. | 14:59.00 |
henrys | marcosw: a debug build and backtrace for 696028? | 14:59.04 |
tor8 | shame about no subdirectories on github then | 14:59.08 |
jogux | tor8: it does. apart from breaking the github mirror until we resolve the subdirs thing :-( | 14:59.14 |
henrys | henrys: from the customer | 14:59.16 |
marcosw | henrys: I'll try to reproduce, I think there is enough information from the customer; if not I'll ask for a sample program. | 14:59.41 |
tor8 | jogux: or, uh, does that make the url a relative to the local repository and not an http url? | 14:59.52 |
jogux | tor8: I don't 100% know how it affects existing checkouts that update, but, even if it doesn't work well, that's only a one off hit at least. | 14:59.54 |
rayjj | marcosw: I will be hoping that you aren't able to reproduce it ;-) | 15:00.14 |
jogux | tor8: I believe it's relative to the upstream git url | 15:00.16 |
| so if you've checked out from git://git.ghostscript.com/mupdf.git, it'll be relative to that. | 15:00.35 |
| tor8: I wondered about the thirdparty ones becoming thirdparty-zlib.git etc? | 15:01.11 |
henrys | marcosw: you have windows 7? | 15:01.12 |
tor8 | jogux: hm, how does that work if you've got multiple remotes? | 15:01.31 |
| jogux: I should just read the manual instead of asking impossible questions :) | 15:02.07 |
jogux | tor8: yeah. I don't know and was about to try and find the manual :) | 15:02.24 |
chrisl | rayjj: I had a look at the problem Yuki raised, but I'm stuck until I get a simulator update - none of my own hackery was able to reproduce the problem | 15:02.39 |
henrys | I'm off to skype now | 15:02.39 |
marcosw | henrys: yes. THat's the only version of windows other than xp that I can halfway understand how to use :-) Though i've played with 10 and it's not as bad as 8 or 8.1. | 15:03.02 |
tor8 | jogux: relative to the 'origin' remote | 15:03.32 |
chrisl | Oh, I didn't around to installing Win10....... | 15:03.41 |
rayjj | chrisl: the doc/Make.htm is woefully out of date. It still mentions ftp://mirror.cs.wisc.edu/pub/mirrors/ghost/GPL/ and I suspect some of the instructions are also 'bit rotted' | 15:03.57 |
henrys | marcosw: that's probably better than me. | 15:04.10 |
tor8 | which might make things messy for robin who, I believe, has his 'origin' pointing to his private casper repo and 'golden' | 15:04.13 |
marcosw | The official win 10 release is tomorrow, I believe. | 15:04.14 |
mvrhel_laptop | yes | 15:04.19 |
tor8 | I set mine up to have 'origin' be the golden repo, and 'tor' is my private | 15:04.29 |
marcosw | my son has been running the preview for a while on his gaming desktop. | 15:04.40 |
rayjj | chrisl: should I open a bug about the doc ? | 15:04.56 |
jogux | tor8: that's how I have things as well. It just means I would need to copies of the submodule repos next to my private mupf repos... I think. | 15:05.00 |
| [I have the same as robin, I mean] | 15:05.12 |
chrisl | rayjj: sure, I'll get to it at some point | 15:05.18 |
tor8 | marcosw: so he didn't get a bust install from microsoft pushing automatic non-optional updates to the graphics drivers that broke with nvidia cards? | 15:05.19 |
mvrhel_laptop | oh that sounds bad | 15:05.47 |
tor8 | jogux: you could just swap the remotes around to be like mine :) | 15:05.47 |
| not being able to turn off updates makes Win10 a no-go in my eyes | 15:06.24 |
jogux | tor8: There's probably more risk of breaking my brain that way :) | 15:06.29 |
marcosw | tor8: he has an nvidia card and hasn't complained that it stopped working, so apparently not. I have a don't ask policy on his machine. | 15:06.42 |
rayjj | chrisl: BTW, is there a way to make a pcl profile build (like the pre-reorg: make pcl-profile) ? | 15:06.59 |
marcosw | I.e. I don't ask him if it has a problem. | 15:07.16 |
chrisl | rayjj: On what platform? | 15:07.36 |
sebras | tor8: weren't you supposed to be able to buy a professional edition where you could control whether or not updates are installed automagically..? | 15:08.00 |
kens | sebras that's enterprise edition | 15:08.12 |
| ad you still haev to install eventually, or it stops updating altogether | 15:08.28 |
rayjj | chrisl: linux (sorry -- was disctracted) | 15:08.58 |
mvrhel_laptop | oh VS 2015 is out now | 15:09.09 |
chrisl | rayjj: make gpcl6pg | 15:09.15 |
| Hmm, maybe I should have dashes in there...... | 15:09.37 |
mvrhel_laptop | I will see if I can get a pro version.... | 15:09.51 |
rayjj | chrisl: there should probably be a list of make targets in the doc somewhere | 15:10.19 |
marcosw | http://gadgets.ndtv.com/apps/news/microsoft-releases-tool-to-stop-automatic-windows-10-updates-720758 | 15:10.29 |
rayjj | (suitably hard to find, of course) ;-) | 15:10.33 |
henrys | chrisl: I always wonder if we should have this proliferation of targets... here just just do XCFLAGS="-O0 -pg" ...but I guess that's cryptic | 15:10.56 |
kens | marcosw that's a pretty swift U-turn :-) | 15:10.59 |
henrys | or -O2 whatever | 15:11.25 |
rayjj | kens: we also want the pg targets in a separate obj and bin directory | 15:11.33 |
tor8 | jogux: bah. creating a repository on github with '/' in it replaces the '/' with a '-' | 15:11.39 |
kens | rayjj maybe that's to chrisl ? | 15:11.46 |
jogux | tor8: :-) | 15:11.47 |
tor8 | but no such friendly transform when cloning | 15:11.49 |
| which would have been helpful | 15:11.53 |
| we should submit a bug report :) | 15:12.07 |
jogux | yeah. ack. | 15:12.09 |
| hehe | 15:12.10 |
chrisl | henrys: the advantage of what we have now is that you can have debug, release, profile, memento builds all available simultaneously | 15:12.22 |
mvrhel_laptop | bbiab | 15:13.00 |
henrys | chrisl: yeah I was going to start using your new mechanism for building outside the tree and stop using the built in directories. | 15:13.12 |
| chrisl: at least the way you explained it I could create a directory "myprofile" or something and just build there, no? | 15:14.29 |
chrisl | henrys: no, because the directories are set at configure time, not make time | 15:14.54 |
henrys | chrisl: okay I think my way would be nicer but no big deal | 15:16.07 |
chrisl | henrys: the other thing is, doing it your way isn't really feasible on Windows, so we'd have even more difference between the platforms :-( | 15:18.17 |
henrys | chrisl: oh didn't realize that. | 15:19.26 |
kens | The current windows solution is preferable to the old one, you can move the solution around at the ghostpdl level and it continues to buid | 15:20.00 |
| Which is great because I can copy the solutoin directory and tehn Git doesn't complain when my VS updates the solution and project files | 15:20.21 |
chrisl | henrys: TBH, whilst I like the convenience of specific targets, I do agree that we seem to have a surprisingly large number | 15:21.27 |
| Most of which are "internal", though | 15:21.56 |
kens | chrisl I thought 696107 should stay with me. | 15:23.26 |
chrisl | kens: I thought we wanted a error thrown? | 15:24.11 |
kens | although I still think its wrong that the TextAlphaBits behaves differtnly to the non-TextAlphaBits, Acrobat handles the file (of course!) so I'll have to hack the PDF interpreter anyway | 15:24.47 |
| I was going to look at changing the graphics library so that it was at least consistent as well though | 15:25.07 |
| Error/non-error just by changing TextAlphaBits is not sensible IMO | 15:25.28 |
chrisl | kens: Am I remember thing wrong? The degenerate matrix one? | 15:25.39 |
kens | chrisl it was that last time yes | 15:25.49 |
| THs time its 'soemthing siliar but different' | 15:26.02 |
chrisl | Well, it's up to you - I don't mind if you take it back | 15:26.18 |
kens | Which my comment to the last one is a predicton of :-) | 15:26.19 |
| Well I'll need to 'fix' the PDF interpreter anyway, just as soon as I finish working out exactly what is causing the matrix to be 'wrong'. I thought it was the 0 Tz but I'm no longer sure. | 15:27.10 |
| I'll take it back | 15:27.17 |
chrisl | Okay - I'm not short of things to do..... | 15:27.35 |
kens | Me neither, but it does look like I should do it. | 15:28.00 |
| Thanks for dealing with the other ones while I was away :-) | 15:28.13 |
chrisl | Wasn't much came up IIRC | 15:28.36 |
kens | Couple of silly ones I htnk, including the one that had been opened on SO then came abd opened a bug report. The crappy output from a Toshiba print driver or something. | 15:29.20 |
| I closed it today :-) | 15:29.30 |
chrisl | Oh, yeh. That was weird..... | 15:29.40 |
kens | I thnk the problem with the text is that the font is deliberatly undefined and then a new one created. | 15:30.06 |
| THe old one is a type 1 the new one is a type 42 | 15:30.14 |
chrisl | Yeh, I saw your comment. Still odd that both we and Adobe get the same output | 15:32.01 |
kens | It is odd, but I don't propose to wrorry about it, especially as we are bug compatible with Adobe :-) | 15:32.35 |
| The driver is totally at fault for the other provlems though | 15:32.50 |
chrisl | Yes, clearly.... | 15:33.10 |
| I have to head out. | 15:33.17 |
kens | bye | 15:33.24 |
| sebras, the bug attached to 696084 is indeed marked private | 15:37.08 |
| I'm sure tor can grab you a copy to look at | 15:37.21 |
henrys | tor8: do you have any help for fredross-perry on his tech at artifex dot com question. fredross-perry has been doing stuff with mupdf and I don't know if he started before you left. | 16:24.31 |
rayjj | chrisl_away: Since you've finished for the day, I will have a go at downloading the sim and duplicating Shinohara-san's problem. I'll email my findings before you start in the AM | 17:02.27 |
| chrisl_away: thanks for getting the sim from him | 17:02.40 |
| chrisl_away: or at least trying. I'll check later to see if he managed to upload the right version | 17:18.17 |
mvrhel_laptop | Robin. did not notice you were here too | 18:21.04 |
| we can chat here if this is better | 18:21.18 |
henrys | fredross-perry: you may want to write tor8 directly sometimes he doesn't see tech messages | 18:40.52 |
fredross-perry | oh, OK. thanks. | 18:41.07 |
mvrhel_laptop | ok bahamas are booked for me | 19:11.36 |
| off to lunch | 19:46.58 |
fredross-perry | @tor please check your email, thanks. | 22:16.44 |
henrys | fredross-perry: I'll catch him in the morning if he doesn't respond. XFA assignment might have temporarily pushed him over the edge ;-) | 22:58.51 |
fredross-perry | not urgent, thanks. | 22:59.07 |
rayjj | grr.. Has anyone tried the current build with VS 2015 Community ? I had it working (at the last meeting) but the changes I did then aren't sufficent now. I am going to uninstall VS2015 RC and reinstall the current, but don't want to duplicate effort if anyone has it working | 23:51.06 |
| Forward 1 day (to 2015/07/29)>>> | |