| <<<Back 1 day (to 2014/01/05) | 2014/01/06 |
kens | Morning chrisl, Happy New Year | 08:05.32 |
chrisl | Happy New Year kens | 08:05.49 |
| Now, what the hell is it I do, again......? | 08:06.08 |
kens | When you're caught up, I have a GC problem I could use some advice with, I'm currently confused..... | 08:06.15 |
chrisl | That's the one that crashes in validate_ptrs()? | 08:06.56 |
kens | I don't recall if that's where it crashesm but its relatively new | 08:07.15 |
| I can see what's happening, I don;'t know why | 08:07.23 |
| Let me know when its convenient to chat about it | 08:07.51 |
chrisl | Okay, I have a couple of e-mails to reply to, and we'll get into it | 08:08.09 |
kens | Then I will go fetch a coffee :-) | 08:08.19 |
| My router is misbehaving, just going to reboot it, back in a moment | 08:23.39 |
| Well back again hopefully.... | 08:41.06 |
chrisl | kens: so this is Bug 694868? | 08:41.29 |
kens | Yes that's the one | 08:41.47 |
chrisl | hrm, I can't reproduce the seg fault :-( | 08:41.58 |
kens | It only exhibits (for me) on a Linux debug build | 08:42.00 |
| In my case 64-bit | 08:42.08 |
chrisl | Yep, 64 bit | 08:42.19 |
kens | Hmm, that's odd | 08:42.27 |
| Can you stick a breakpoint in the 'RELOC_PTRS_WITH' macro at the head of gdevpdf.c ? | 08:43.27 |
| The line that says 'gx_device_pdf_do_ptrs(r1) (whichj is itself a macro...) | 08:43.34 |
chrisl | God, why have bonkers, single use macros, that's a pile of crap! | 08:46.43 |
kens | Yes, this is part of my problem, the pdfwrite stuff defines its only insane set of macros :-( | 08:47.09 |
| s/only/own/ | 08:47.17 |
| What command line are you using by the way ? | 08:47.39 |
chrisl | The one in the bug | 08:48.13 |
kens | OK same as me then | 08:48.20 |
| Did you hit that breakpoint ? | 08:48.30 |
chrisl | Yes | 08:48.33 |
kens | OK what happens for me is that pdev->outline_levels gets relocated | 08:48.49 |
| WHich is fair enough, its one of the enumerated pointers | 08:48.59 |
| But.... After its relocated, its content is different | 08:49.10 |
| Before relocation its all 0x00, afterwards there is random data there | 08:49.27 |
| DO you see the relocation ? | 08:49.39 |
| (I say 'random', but actually its always the same) | 08:50.13 |
chrisl | Debugger just crashed.... | 08:51.28 |
kens | ah that doesn't help | 08:51.35 |
| I managed to find out how to get ddd to be conform to my notion of when a dialog should be available to type in :-) | 08:52.21 |
chrisl | To type what in? | 08:52.59 |
kens | things like the variables when displaying variables | 08:53.13 |
chrisl | Like the gdb console? | 08:53.35 |
kens | Umm, not sure, I've not used gdb for so long | 08:53.46 |
| When you display a variable, ddd pops up a little dialog for you to enter the variable name | 08:54.07 |
| I;'ve alwsays had trouble typing in it, even when it apparently has the caret. | 08:54.24 |
chrisl | Oh, I usually just click it in source display, right click and use the context menu | 08:55.23 |
kens | Turns out the 'keyboard focus' was set to 'point to type' instead of 'click to type' even when the cursor is merrily blinking in there, unless you move the mouse to it, you can't type there. I find this hard to deal with.... | 08:55.25 |
| In this case, the variable isn't in the source display :-( | 08:55.53 |
| So if you can't reproduce this crash I'm kind of stumped. | 08:56.21 |
chrisl | Yeh, I *think* that's actually a lesstif bug, rather than a ddd bug - it should pick up the click/point focus from the window manager | 08:56.29 |
| Debugger crashed again - this ain't good :-( | 08:56.54 |
kens | O.O | 08:57.00 |
| At least that's not happening for me | 08:57.09 |
chrisl | I thought you were going to say "at least it's consistent"! | 08:57.30 |
kens | Consistently crashing the debugger isn't good :-( | 08:57.49 |
| Its consistent for me here too though | 08:58.03 |
chrisl | No, but it's better than randomly crashing the debugger....... | 08:58.16 |
kens | Well I guess consistent crashes are better than random ones, yes .... | 08:58.45 |
| I do wonder if its the same problem though, if its crashing the debugger in the same place all the time | 08:59.01 |
chrisl | Could be - it's hard to tell, tbh | 08:59.53 |
kens | OK well how about looking at the code and telling me if I've done something dumb ? | 09:00.15 |
chrisl | Where does pdev->outline_levels get relocated? | 09:01.12 |
kens | icky | 09:01.20 |
| The relocaTION IS IN 'GX_DEVICE_PDF_DO_PTRS | 09:01.34 |
| sorry that should be lower case | 09:01.42 |
| the macro is defined in gdevpdfx.h | 09:01.53 |
| which just expands each of the pointers to be called with RELOC_PTR | 09:02.24 |
| Notice in RELOC_PTRS_WITH we define 'r1' as RELOC_PTR(gx_device_pdf,elt) and tehn keep on redefining it later | 09:03.05 |
chrisl | Which I assume is because it is, itself, used as the parameter to a macro? | 09:03.47 |
kens | pdev->outline_levels is allocated in pdf_open() at about line 711 | 09:04.00 |
| chrisl I assume that's why its done that way, the whole thing is horrendous | 09:04.13 |
| THe line following the allocation sets the memory contents to 0 | 09:04.47 |
| And since we have no outlines, that's how it should remain throughout | 09:05.00 |
chrisl | I must admit, I can't see the actual reason for the r1 macro..... | 09:05.10 |
kens | I suspect this is more cargo cult programming | 09:05.31 |
| Probably copied from elsewhere | 09:05.38 |
| it looks like r1 is intended as a mnemonic for 'relocate 1' or something. THere's a 'e1' defined above in the enumeration code | 09:07.26 |
| The r1 macro does also drop the 'index' of the pointer, so it clearly has no value apart from making it easier to count the number of pointers (which I think we no longer need ;-) | 09:08.00 |
| Ah OK the reason for the macros r1 and e1 is because we reuse the 'gx_device_pdf_do_ptrs' macro for both enumeration and relocation. | 09:09.50 |
| I still think all this macro usage is just confusing though | 09:10.09 |
chrisl | I'm still not convinced it's a good idea..... | 09:10.21 |
kens | I think its harmless, just very hard to understand | 09:10.36 |
chrisl | Do you get the outline_levels relocated on the first call? | 09:10.54 |
kens | yes | 09:11.00 |
chrisl | Ah, I don't..... | 09:11.09 |
kens | I don't on Windows, which is probably why it works there | 09:11.27 |
| I get a relocation at the end of initialisation due to a gs_vmreclaim | 09:12.05 |
| I wonder what would affect the relocation.... | 09:12.44 |
| This VM only has 2 Gb VM | 09:13.26 |
| ('only' ROFL...) | 09:13.55 |
chrisl | Physical memory is unlikely to affect it. But if we're right on the edge of the VM trigger threshold, a tiny change could tip it - so glibc/gcc versions could be in play | 09:14.27 |
kens | I'm pretty sure my version is quite old, just a second | 09:14.45 |
| 4.5.1 20100924 | 09:15.12 |
chrisl | 4.8.1 here | 09:15.36 |
kens | Hmm, I could upgrade I guess, but then if the problem goes away I can't even reproduce Macros's result | 09:16.09 |
chrisl | No, I'd keep what you've got. If you feel the need for something more up to date, make a new VM | 09:16.43 |
kens | Yes, true, I have the disk space. | 09:17.04 |
| OK so where to go from here ? I can't figure out the GC stuff, it seems to me it should not alter the contents of the memory like that when relocating. | 09:17.52 |
chrisl | Have you tried it with -Z@ ? | 09:18.16 |
kens | I could break out the pointer individually so I could at least follow it into the memory manager | 09:18.17 |
| No, haven't tried -Z@ what does that do ? | 09:18.26 |
chrisl | It fills memory with bit patterns, so you can see if it thinks the memory is freshly allocated, freed or gc'ed | 09:19.05 |
kens | Seg faulted with no output | 09:19.08 |
| OK then let me start ddd again | 09:19.19 |
| Well the memory contents are different, but similar | 09:21.17 |
| The 0x0c00 are the same, the 0x00 have changed to 0xa1 | 09:21.37 |
chrisl | I think 0xa1 is freshly allocated memory - so that is odd..... | 09:22.11 |
kens | That's after the relocation of course | 09:22.17 |
| It looks like the memory is 'partially' filled | 09:22.30 |
chrisl | kens: give me an hour or so to play with this, and I'll see what I can find | 09:23.45 |
kens | Yes, comparing with the run without -Z@ the 0xa1 replaces memory that was 0x00, the non 0x00 bytes are the same | 09:24.00 |
| chrisl OK I'll keep poking it too | 09:24.09 |
chrisl | kens: you seem to be using a different allocator for outline_levels compared to the other objects in the structure - I'm suspicious about that...... | 09:54.15 |
kens | Am I ? THat's odd. | 09:54.26 |
| Just a moment | 09:54.37 |
| Hmm I see | 09:55.08 |
chrisl | I suspect you're using normal gc'ed memory (subject to save/restore) the other stuff is using stable_memory (gc'ed, but not subject to save/restore) | 09:55.15 |
kens | mem is gs_memory_stable(pdev->memory) | 09:55.20 |
| Definitely don't want it subeject to save/resore, that won't work well | 09:55.44 |
| let me try changing that | 09:56.13 |
chrisl | Well, pdev->memory != pdev->memory->stable_memory (in this case) | 09:56.27 |
kens | THat certainly solves the debug problem | 09:58.05 |
| let me just run a cluster test | 09:58.11 |
| Just remembered, I need to check where I reallocate the memory too.... | 10:00.51 |
| Hmm, I wonder if this could be the source of other problems.... | 10:02.43 |
| Ah OK pdev->pdf_memory is what I should be using, its always the stable_memory | 10:03.19 |
chrisl | Which does rather make sense...... | 10:04.34 |
kens | Yes, I was just wondering if some of the other weird GC bugs are the ame problem. | 10:04.56 |
| And yes, there was the same problem in the reallocation, fixed that too. | 10:06.17 |
| I should review the other memory uses, but a quick look in pdf_close and they look OK, mostly at least | 10:06.35 |
| I think this must be the root of the problem, and I'm kind of surprised a regular cluster test didn't show it up. Thanks for the help chrisl. | 10:07.36 |
chrisl | kens: no probs. It's possible that just dumb luck meant that in the optimised builds we simply ignored the problems - optimised builds don't do the validate_object() stuff. | 10:10.46 |
kens | The surprise to me is that the free was using the stable memory allocator, and the allocation was done using the non-stabl;e allocator. I woudl have expected that to throw a problem..... | 10:12.26 |
chrisl | Again, I think we'll just ignore that in an optimised build, and possibly just issue a warning in a debug build *if* we get there | 10:13.50 |
kens | Hmm well we got there in the WIndows debug build, without any warnings. | 10:14.16 |
| I guess its OK, just something I had overlooked, thanks again | 10:14.38 |
chrisl | kens: yeh, it looks like when freeing, if we can't find the chunk containing the address we're trying to free in the supplied allocator, we'll just silently ignore it, unless we're running with a debug switch - which one, I'm not sure...... maybe we should change that | 10:19.22 |
kens | aYeah, at least for debug builds we should always warn I think | 10:19.57 |
| It seems pretty indicative of a fatal problem | 10:20.10 |
| dAt least it explains why I didn't see any problems... | 10:21.08 |
chrisl | Yeh, I've stumbled on this a few times before. I think I will change it so it warns in debug builds. | 10:21.17 |
kens | Great, thanks! | 10:21.31 |
chrisl | kens: btw, I get a warning about redefinition of "PDF_FOR_EPS2WRITE" because it's not being undefed after each #include "gdevpdfb.h" | 10:49.17 |
kens | wYeah I should probably fix that, its harmless though | 10:49.37 |
chrisl | Hmm, having the memory warning means quite a performance hit :-( | 10:52.48 |
kens | Well, if its only on a debug build.... | 10:53.07 |
| Oh but Marcos won't be happy | 10:53.21 |
| OK forget it I guess | 10:53.53 |
chrisl | If we can remember, we can check with marcos when he's around. | 10:55.00 |
kens | If it hits the performance too much he won't get the nightly runs done. | 10:55.22 |
| I'll probably remember this now, since I've been bitten by it... | 10:55.38 |
chrisl | I'm trying to work out how to get the warning to appear - it's not obvious what combination of flags to use :-( | 10:56.31 |
kens | wHmm, that's not too helpful... | 10:56.58 |
chrisl | I think it's supposed to happen with -Z? | 11:00.39 |
kens | Well, if you haven't updated, the you can test with the bug :-) | 11:00.59 |
chrisl | Yes, it helped when I used pdfwrite! So, yeh, it's -Z? and you get "outline_levels array: freeing 0x1dd8158, not owned by memory 0x1a67568!" | 11:02.33 |
kens | OK well tht's helpful, I'll try and remember the invocation | 11:03.01 |
paulgardiner | Robin_Watts: two commits on paul/master - second one at last makes scrolling smooth on my S2. | 12:35.29 |
Robin_Watts | Cool. | 12:35.37 |
| paulgardiner: Both look plausible to me. | 12:52.06 |
paulgardiner | Ta | 12:52.11 |
Robin_Watts | paulgardiner: 2 commits on robin/master (Well, 4, but ignore the top 2) | 13:18.42 |
paulgardiner | Okay | 13:18.50 |
Robin_Watts | The 'document handlers' one was something I discussed with tor. We should get his opinion on it too. And sebras if he's about. | 13:19.10 |
paulgardiner | First one: should we worry about the possibility of avail_out being negative? | 13:39.10 |
| ... passing a negative number to memset | 13:40.09 |
| I haven't looked what type of int avail_out is. If signed, you should probably check > 0, rather than != 0. | 13:43.35 |
kens | chrisl did an audit of all the places in pdfwrite that use pdev->memory instgead of pdev->pdf_memory, think I've found another mismatch. | 13:54.42 |
chrisl | kens: Oh, good. Although, if it's something used often, I'm surprised it hasn't thrown a seg fault in *some* file | 13:57.18 |
Robin_Watts | paulgardiner: availOut = uInt | 14:00.03 |
paulgardiner | Oh okay | 14:00.16 |
| Robin_Watts: perhaps fz_register_document_handlers should be fz_register_default_document_handlers. | 14:01.46 |
Robin_Watts | paulgardiner: perhaps. I had 'all' for a while, but I felt the name was long enough already :) | 14:02.18 |
kens | chrisl its the font name of the base font. I'm guessing we don't (often) restore those away before the end of the job | 14:02.27 |
chrisl | kens: possibly, I can't honestly remember..... | 14:03.09 |
kens | Me neither, but it looks pretty clearly wrong. All the other places seemed 'OK' | 14:03.30 |
paulgardiner | Robin_Watts: looks good to me. See what the others think. | 14:06.52 |
Robin_Watts | hey ray_laptop. | 15:10.06 |
| You may have answered this (but I haven't seen it in the logs), or maybe you're just sick of being asked, but how is the recovery going? | 15:10.42 |
ray_laptop | Hi, Robin_Watts. Still not much difference. I'm getting used to keeping my eye lubricated, so the discomfort is less, but I still can only have my left contact in for about 1 hour or so, so I can't really help with all of the errands | 15:15.21 |
| I still have really strange looking smile | 15:15.48 |
chrisl | ray_laptop: have you had any medication to help? | 15:17.19 |
Robin_Watts | not being able to help with all the errands sounds like an advantage :) | 15:18.35 |
ray_laptop | chrisl: I was on Acyclovir for the first 10 days to help get rid of the shingles (maybe a bit sooner), and steroids for 4 days. On the 31'st I went back and since there was no improvement, they gave me another round of steroids at a slightly higer dose, for 7 days | 15:19.19 |
| Robin_Watts: I get a bit stir crazy, particularly since the kids have been home and have needed attention. | 15:20.06 |
Robin_Watts | yeah, I can imagine. | 15:20.15 |
chrisl | ray_laptop: Right, it's a steroid they prescribe over here. Unfortunately, it seems to be one of those ailments that just has to run its course :-( | 15:21.26 |
ray_laptop | Robin_Watts: but I take a couple of walks a day for 20 minutes or so. I _was_ doing errands with my Dad driving, but they've gone back to STL now. They have 2F for the *high* today -7F now. They're thinkg that they should have stayed with unitl spring | 15:22.13 |
| chrisl: yeah. I know one person that got over it in 5 days, and another that was 3 months | 15:22.47 |
Robin_Watts | ray_laptop: Yeah, the mid->east of the US is getting really hammered by cold at the moment, right? | 15:24.03 |
kens | *sever* cold | 15:24.13 |
Robin_Watts | Supposedly that's supposed to hit us later this week. | 15:24.20 |
| kens: freezing your * off ? | 15:24.31 |
chrisl | ray_laptop: I don't know if it's worth taking a course of over the counter anti-inflammatories once the steroids have run out...... | 15:24.59 |
kens | I meant the US | 15:25.00 |
henrys | yup single digits here. | 15:25.05 |
kens | Oh good, if it hits us then all the lnew ponds can freeze..... | 15:25.37 |
henrys | mvrhel_laptop: how's family in michigan I've heard really bad weather reports? | 15:25.57 |
Robin_Watts | The only place that's not got snow is mvrhel_laptop's local ski run, right? :) | 15:25.58 |
ray_laptop | Robin_Watts: our ski mountains as well. Just man-made | 15:33.57 |
henrys | we have plenty of snow here but everyone is stoned so I don't know if it is safe. | 15:35.43 |
ray_laptop | mvrhel_laptop: I found a file: tests_private/pdf/sumatra/x_-_blend_modes.pdf that paints to the page with different BM's. It seems that gs_setblendmode just modifies the pgs, and doesn't actually push any compositor action (such as PDF14_SET_BLEND_PARAMS) | 15:37.29 |
| henrys: it's probably safe early in the morning when the stoners are still sleeping | 15:38.52 |
mvrhel_laptop | good morning | 16:26.53 |
Robin_Watts | Hi tor8. | 16:26.57 |
| Morning mvrhel_laptop | 16:27.15 |
tor8 | Robin_Watts: hi. you used the wrong number (I'm switching that phone off permanently in a month or so) | 16:27.18 |
Robin_Watts | tor8: oh, sorry. Let me try and change to your new number. | 16:27.38 |
mvrhel_laptop | we did finally get a bit of snow here on the slopes at one of the ski resorts. half way decent. the place close to me is still closed :( | 16:27.58 |
| henrys: yes, the temps are really low there in michigan | 16:28.19 |
| ray_laptop: if you can open a bug for me about the blend modes, that would be great | 16:28.47 |
henrys | mvrhel_laptop: sabrina's sister has 3 feet but they're out in the boonies | 16:29.03 |
mvrhel_laptop | wow | 16:29.48 |
Robin_Watts | tor8: OK, number updated. | 16:31.01 |
| tor8: So, various small things and 1 thing to discuss. | 16:31.10 |
| firstly, there are 2 commits on robin/master (actually 4, but ignore the top 2). | 16:31.37 |
| paulgardiner has looked at them, but at least the document handlers one deserves a wider look. | 16:32.02 |
ray_laptop | mvrhel_laptop: I'm not sure that it's a bug. For one thing, it call gs_setblenddmode BEFORE it pushes the pdf14 device. | 16:32.05 |
Robin_Watts | I've gone back to the JNI stuff and started to look at that again. | 16:32.32 |
tor8 | Robin_Watts: truncated zip/png fix looks good | 16:32.45 |
Robin_Watts | I think the fitz wrapping is looking OK. | 16:32.54 |
| I've just started on the pdf wrapping. | 16:33.02 |
| and to make pdf objects, we need a pdf_document *. | 16:33.27 |
| Currently, if we use the fz_document interface to open stuff, we have no way of getting a pdf_document * other than by doing a direct cast. | 16:33.54 |
| I was wondering if we should standardise on some sort of fz_document call for upcasting. Or downcasting. Whatever. | 16:34.36 |
| Perhaps fz_document should have a pointer to the document handler in use? | 16:35.11 |
| That way we can check for doc->handler == pdf_document_handler, and only cast to a pdf_document * if that's true or something ? | 16:35.45 |
tor8 | document handlers look good, but I wonder about the spelling of recogni[sz]e? we use US spelling elsewhere. | 16:36.02 |
| Robin_Watts: didn't paul add a function pdf_specifics() just for that? | 16:37.06 |
Robin_Watts | You're right. It should be recognize I guess. | 16:37.10 |
| tor8: So he did! | 16:37.29 |
| I can live with that for now then. | 16:37.36 |
tor8 | up/down-casting calls might be handy, but I don't think we'll use it for anything but PDF for the foreseeable future | 16:38.34 |
Robin_Watts | If I fix 'recognize' are you happy with that commit then? | 16:38.37 |
tor8 | Robin_Watts: yes. | 16:38.45 |
Robin_Watts | We can do a more generic version when we need it. | 16:38.47 |
| Brilliant. I shall fix that and push it then. thanks. | 16:38.57 |
| This leaves us with just 1 regularly indeterminate file, the non_employee_travel thing in the mujstests. | 16:39.28 |
tor8 | I'm not happy about the no-run split but we still need that to keep the bloat down from stuff that doesn't need embedded fonts/cmaps | 16:39.34 |
Robin_Watts | I should try to figure out what that's about at some point. | 16:39.54 |
| tor8: I'm happier with the no_run split now we have doc_handlers :) | 16:40.42 |
tor8 | Robin_Watts: I will traveling part of the day tomorrow so may not be reliably available (and today was a holiday here, but no sweat) | 16:41.00 |
Robin_Watts | tor8: ah, sorry. | 16:41.16 |
tor8 | don't worry, I'm just playing computer games anyway :) | 16:41.30 |
Robin_Watts | tor8: Did you survive the steam sale with your wallet intact ? | 16:42.20 |
tor8 | the steam sale was easy, it was the gog sale that ruined me :( | 16:42.32 |
Robin_Watts | I only bought 1 thing from gog, and that was very cheap. | 16:43.06 |
tor8 | did you get the old fallout games for free during the giveaway? | 16:43.20 |
Robin_Watts | I did. | 16:43.25 |
| And I carefully restricted my spending on the steam sale so I only bought daily deals or flash deals, so I didn't have the "gah, I bought it yesterday, and it's half the price now" thing that's hit me before. | 16:44.26 |
tor8 | yeah. should wait until the last day to buy the stuff that you didn't see on the flash sales :) | 16:44.56 |
Robin_Watts | Arkham origins was in the first days sale (and I missed it then) and was reprised on the last days sale, so I grabbed it then. | 16:45.36 |
tor8 | I picked up a handful of games you mentioned on the gog sale, gateways and the adventures of shuggy | 16:47.01 |
| oh, and I played through most of the new XCOM Enemy Within expansion. sadly, it got very same-y towards the end so I ran out of steam before finishing it | 16:48.35 |
KermitTheFragger | hi all im trying to compile ghostscript 8.71 (for gsdjvu) on a Solaris derative (OmniOS) without X, but '--without-x' does not seem to do the trick? | 16:49.01 |
tor8 | if they'd reduced the number of filler missions you have to do, it would've been a perfect game | 16:49.03 |
kens | Thar's a shame, I enjoyed trhe original X-COM | 16:49.08 |
KermitTheFragger | tor8: may i recommend: http://openxcom.org/ instead ? :-) | 16:49.15 |
tor8 | KermitTheFragger: which one is that? there's at least several different open source clones of the original xcoms :) | 16:49.51 |
KermitTheFragger | tor8: the best :-) its basically like openttd, new engine, same gameplay | 16:50.09 |
chrisl | KermitTheFragger: 8.71 is rather old, although I can't recall having changed the X related stuff, I may well have..... | 16:50.45 |
tor8 | KermitTheFragger: I shall have to look into that. | 16:50.47 |
KermitTheFragger | yeah i know its old :-( but gsdjvu doesn't work with 9 | 16:51.14 |
tor8 | KermitTheFragger: maybe it's --disable rather than --without? I can't remember how well the autoconf stuff actually works with the various options. | 16:51.22 |
KermitTheFragger | tor8: the enable option is '--with-x' | 16:52.18 |
| ill give it a shot and see what it does | 16:52.40 |
chrisl | KermitTheFragger: --without-x removes the X display devices when I try it. What error(s) do you get? | 16:54.22 |
KermitTheFragger | chrisl: it chokes on: In file included from ./base/gdevxalt.c | 16:55.24 |
| http://pastebin.com/fEfaNMhG | 16:55.56 |
chrisl | Hm, I've no idea what that is for :-( | 16:57.15 |
| But it definitely shouldn't be getting built..... | 16:59.41 |
KermitTheFragger | im trying with -without-x11 and --disable-x11, see if that wields anything | 17:01.46 |
ray_laptop | mvrhel_laptop: I found one thhing that was confusing me. The pdf14_opcode_names never had the PDF14_ABORT_DEVICE inserted in the list | 17:01.53 |
| so the -Zv output was confusing me a bit | 17:02.07 |
chrisl | KermitTheFragger: can you pastbin the Makefile that you get from configure? | 17:02.45 |
KermitTheFragger | chrisl: sure | 17:02.54 |
| chrisl: http://pastebin.com/9VgTUQEt | 17:04.17 |
mvrhel_laptop | having cable modem issues here that I am trying to get resolved with cable modem provider so I may be off and on a bit | 17:04.46 |
KermitTheFragger | thats with the --without-x11 and --disable-x11, not with --without-x (don't even know if those 2 are valid options) | 17:04.51 |
| but --without-x wields the same result | 17:05.09 |
chrisl | KermitTheFragger: DEVICE_DEVS section looks *very* different from the Makefile I get out of configure.... | 17:06.53 |
KermitTheFragger | chrisl: hmmm could be that the gsdjvu voodoo messes it up | 17:07.09 |
chrisl | KermitTheFragger: I've no idea what they add/remove change. But using our 8.71 package, configure --without-x causes *none* of the X devices to be built. | 17:08.16 |
KermitTheFragger | chrisl: thanks, guess I was barking up to wrong tree then, could have figured that out myself, sorry for that | 17:09.40 |
chrisl | KermitTheFragger: no worries, you can still find our 8.71 distribution here: http://downloads.ghostscript.com/public/ | 17:10.51 |
| You could give that a try, and compare the results | 17:11.03 |
KermitTheFragger | will do! thanks! | 17:11.31 |
kens | TIme to go, goodnight everyone | 17:34.45 |
Robin_Watts | night kens | 17:34.50 |
henrys | does anyone know where marcos' documentation for "custom" testing is? I know I got an announcement about it, now I can't find it. | 17:59.07 |
Robin_Watts | tests_private/comparefiles/custom_tests.lst | 18:08.15 |
| Mail from Marcos to tech, 8 Nov 2013 04:46 entitled "Custom regression tests" | 18:08.40 |
| henrys^ | 18:09.06 |
henrys | Robin_Watts: thanks no idea how I couldn't find that. | 18:10.08 |
Robin_Watts | thunderbird failed to find it with a search for me :( | 18:10.36 |
| mvrhel_laptop: ping | 18:15.25 |
| Marti has been continuing to bash on the lcms2 threading fixes. | 18:15.50 |
| With the mechanism he's picked, he is still pickling the context into the transforms and the links. | 18:16.15 |
| This means that we can't use a single transform or link from more than one thread at a time. | 18:16.37 |
| I don't think this is a problem for us, is it ? | 18:16.45 |
mvrhel_laptop | Robin_Watts: that is the case that we have currently. | 18:22.45 |
Robin_Watts | Would it be a significant advantage for us to be able to avoid that? | 18:23.16 |
mvrhel_laptop | It would be nice to be able to use them in multiple threads | 18:23.24 |
| I had hoped that was the case when I started this project years ago | 18:23.44 |
Robin_Watts | OK. I'll mention this to Marti and see if it's an easy thing to lift. | 18:23.44 |
| I suspect that it will involve him passing cmsContext *'s in more places. | 18:24.01 |
| and in particular it will involve the function pointer type for the pipeline stages changing. | 18:24.18 |
| which will require changes in all plugin code. | 18:24.36 |
| so I suspect he might not be keen on that. | 18:24.43 |
| but I'll mention it. | 18:24.51 |
mvrhel_laptop | ok. Thanks Robin_Watts | 18:33.29 |
ray_laptop | might be nice to be able to collect the "lock" time for the CMS with NumRenderingThreads set to 4 or so. In patricular, on Windows this is an issue since locks are such a performance hit | 19:12.31 |
henrys | marcosw:I thought http://bugs.ghostscript.com/show_bug.cgi?id=694837 would be good for the testing, but no double conversions. What do you recommend to test it? | 19:13.49 |
| marcosw: I assume there is a nightly that does luratech jbig2 encoding | 19:14.27 |
| marcosw: good for custom_tests.lst that is. | 19:16.37 |
florian_pb | hello | 22:09.13 |
ghostbot | privet | 22:09.13 |
florian_pb | I use pdf2ps to convert a PDF to Postscript. The PDF was created using ghostscript before using params sProcessColorModel=DeviceGray and sColorConversionStrategy=Gray. Thus all images in the PDF are converted to grayscale and if I analyse the file with ghostscripts inkcov device only the K component is used. After passing the PDF to pdf2ps the resulting Postscript file is still in grayscale | 22:15.15 |
| but more CMYK components, not only K, are used. Is this a desired behaviour? | 22:15.15 |
ray_laptop | florian_pb: We'd probably have to see the PDF file you created, however, pdf2ps generally uses RGB or Gray so I wouldn't expect any CMYK colors if the original PDF didn't have those colors. | 22:41.49 |
| florian_pb: please open a bug, attach the PDF file, and MAKE SURE and state the version (and distro) for the GS you are using and the pdf2ps command line | 22:42.45 |
raddude | bots | 22:43.53 |
| hi | 22:44.07 |
ray_laptop | Hi, raddude | 22:44.18 |
raddude | in the process of setting up a pdf viewer extension for my mediawiki... and it requires ghostscript | 22:45.10 |
| just joined this IRC to see if anyone has done this previously... i assume it's not very likely | 22:45.51 |
ray_laptop | raddude: OK. Lots of distributions have GS | 22:45.56 |
| raddude: actually, there are a lot of web sites that use GS to convert PDF to images (for back when most browsers couldn't handle PDF) | 22:46.44 |
| raddude: but I don't mess with setting up wiki's | 22:47.17 |
| but if your mediawiki says it requires ghostscript, then just install it | 22:47.52 |
| we recommend the latest release (9.10) of course | 22:48.12 |
raddude | gotcha. i'll go ahead and start setting it up | 22:48.42 |
| just curious, what is this IRC used for? | 22:49.09 |
ray_laptop | raddude: what OS / distro are you on ? | 22:49.16 |
raddude | mediawiki is running on MS server 2008 | 22:49.59 |
florian_pb | ray, thank you. | 22:50.01 |
ray_laptop | raddude: conversations about Ghostscript and mupdf (Artifex Software's primary products). Most of the Artifex staff are on during their workdays and many of the users and contributors to either product hang out here | 22:50.22 |
| raddude: OK. Well, there are 32 and 64 bit ghostscripts for Windows on the download site | 22:51.01 |
| and we know it runs on Windows Server versions (some of our licensed customers use it) | 22:51.32 |
raddude | cool, i'll try out the x64 version | 22:51.54 |
| thanks ray | 23:00.28 |
| adios | 23:00.29 |
ray_laptop | raddude: stop back if you have problems | 23:01.02 |
raddude | will do! | 23:01.15 |
ray_laptop | HEY!! What happened to my git repo ??? | 23:11.54 |
| on http://git.ghostscript.com/ it skips from paulg/mudf.git to robin/gostpdl/git | 23:12.48 |
| Is there something that someone didn't tell me about my tenure at Artifex :-( | 23:13.14 |
| Robin_Watts: marcosw: Who maintains this ??? | 23:13.43 |
| mvrhel_laptop: Robin_Watts: I have a patch I'd like to throw out for review. My users/ray/ghostpdl.git is gone -- what's the best way to have it reviewed ??? (or should I bother) | 23:15.04 |
mvrhel_laptop | ray_laptop: you can always email me the patch | 23:15.39 |
ray_laptop | This finally enables the pdf14 clist optimization that somehow got dropped on the floor. I hadn't seen any differences because of one patch missing in gxclread.c, so I thought it was done | 23:16.09 |
mvrhel_laptop | oops | 23:16.25 |
ray_laptop | mvrhel_laptop: OK. I'll send it to you and Robin | 23:16.26 |
| mvrhel_laptop: Robin_Watts: email requesting code review sent. Let me know. | 23:37.58 |
| Robin_Watts: BTW, I _do_ want to know what happened to my git repo on git.ghostscript.com/user/ray | 23:38.36 |
Robin_Watts | ray_laptop: Hi | 23:58.30 |
| Forward 1 day (to 2014/01/07)>>> | |