| <<<Back 1 day (to 2020/10/19) | Fwd 1 day (to 2020/10/21)>>> | 20201020 |
malc_ | recursive cloning github's mupdf repository eventually leads to - Cloning into '/home/malc/xsrc/llpp/build-test/mupdf/thirdparty/leptonica'... | 00:07.14 |
| Username for 'https://github.com': | 00:07.14 |
| | 00:07.15 |
| and this fails non-interactive scripts | 00:07.57 |
user38 | Greetings everyone. | 05:09.42 |
sebras | user38: welcome back. | 05:51.25 |
user38 | Thank you kindly. Per https://bugs.ghostscript.com/show_bug.cgi?id=703004#c7, I'm preparing 1.18.0 for building. | 05:52.08 |
sebras | user38: ok. | 05:52.37 |
malc_ | is excited and holds his breath... really | 06:16.14 |
| great | 06:37.42 |
user38 | I got kicked out with "invalid captcha". No idea why. Anyway, the build blew up thus: | 08:18.25 |
| > gmake CC=gcc CXX=g++ | 08:18.48 |
| LINK build/release/scripts/hexdump.exe | 08:18.51 |
| ld: fatal: unrecognized option '--' | 08:19.01 |
| gmake: *** [Makefile:76: build/release/scripts/hexdump.exe] Error 1 | 08:19.16 |
| looks like a GNU ld(1) hard-coded --long-option in the Makefile, have to investigate further. | 08:19.46 |
malc_ | user38: you are on a mac with llld? | 08:20.45 |
user38 | Solaris 10. | 08:22.44 |
| AT&T SVR4 ld(1). | 08:22.56 |
malc_ | ah so surs linker | 08:22.57 |
| user38: "gmake [..] verbose=yes" can help | 08:23.26 |
user38 | Ah! The rejected patch chunk is a clue: | 08:25.22 |
| LDFLAGS += -g -rdynamic | 08:26.08 |
| -rdynamic is a linker-specific option. | 08:26.37 |
| And every linker impements it differently. | 08:26.52 |
| s/impements/implements/ | 08:26.58 |
| -pg also. | 08:27.22 |
| I'll have to look at this closer before I can proceed. | 08:27.39 |
| ifeq ($(OS),MACOS) -- where is "MACOS" set, respectively if I do the following, will it work: | 08:36.21 |
| ifeq ($(OS),SUNOS) <- will that work? | 08:36.46 |
| Ah, I found it, it's just uname(1). So the string should be "SunOS", not "SUNOS"... | 08:37.51 |
Robin_Watts | user38: So, did that allow you to build 1.18? | 11:25.27 |
user38 | I had to take a detour down the rabbit hole... working my way upwards now. | 11:49.11 |
| It all started with the UltraSPARC T1 being too slow, and snowballed from there. | 11:49.42 |
| Was trying to put together patches and the source in a coherent whole, xz(1) on an UltraSPARC T1 is an exercise in patience. | 11:50.39 |
ator | xz is an exercise in patience on *any* machine... | 11:54.24 |
user38 | But the compression it squeezes out of files is unbeatable. I found a way to speed it up for transient work: xz -9e -T0 --block-size=1m ... but that sacrifices compressed file's size significantly. | 11:59.28 |
| Okay, all the patches I made for 1.17.0 apply cleanly to 1.18.0 now; compiling... | 12:01.44 |
| Does Artifex have any Solaris systems in-house? | 12:02.21 |
chrisl | user38: "Artifex" does not. I have my own Solaris/SPARC box, but it's very old and slow | 12:06.14 |
user38 | So how do you get it to compile on Solaris, where the Linux timegm(3) function does not exist? | 12:07.04 |
chrisl | I don't work on mupdf.... I work on ghostscript | 12:07.24 |
user38 | That explains that then. | 12:08.26 |
chrisl | And Oracle are not very helpful where the more modern SPARC hardware and Solaris environment are concerned | 12:10.36 |
user38 | That's because they're Snoracle. | 12:10.55 |
| Also Solaris 11 changed way too many things way too radically, so most of us Solaris people abandoned it for a some illumos distribution variant. Solaris 10 is the last good Solaris. | 12:12.06 |
| s/a some/some/g | 12:12.18 |
chrisl | I have 10 and 9 in a container. Actually, my Solaris machine is probably not much slower than your T1 powered one | 12:13.41 |
| user38: If the latest patches you're testing still have problems, I'll crank the handle on my machine, and work with Robin_Watts on getting the wrinkles ironed out. | 12:15.17 |
Robin_Watts | user38: If you have patches to improve buildability on Solaris, please let us know! | 12:16.30 |
user38 | I was going to discuss this with you after this action is successfully closed, how you would like those... in another bug report or by some other venue perhaps? | 12:17.27 |
Robin_Watts | A bug report is probably the simplest, cos it means that all of us can see them. | 12:19.00 |
Robin_Watts | grabs lunch. bbs. | 12:19.29 |
user38 | I would think so too, although in the interest of being completely open and fair, I'm telling you in advance, be prepared to cherry-pick the Solaris fixes from my patches... When I embarked on porting MuPDF to Solaris, I knew literally nothing about the build engine, so my patches are written based on my reverse engineering of the MuPDF's build | 12:21.34 |
| system, and designed to get me over the line building and working on Solaris. | 12:21.35 |
Robin_Watts | lunch denied :( | 12:23.08 |
| user38: Sure. | 12:23.17 |
user38 | Okay, 1.18.0 built! | 12:24.21 |
| Now let's see how we fare... | 12:24.29 |
| So, 32-bit, same as before: a simple test PDF without graphics renders in a window without decorations(!), a more complex PDF with graphics makes it core dump with a bus error, same as before. | 12:27.17 |
| Now on to building 64-bit... | 12:27.24 |
Robin_Watts | ok, so once you've done with 64bit, if you could do a 32bit debug build, run that, and catch it in a debugger when it dies. | 12:34.10 |
| ator: https://git.ghostscript.com/?p=user/robin/mupdf.git;a=commitdiff;h=6292d6d3b20ab450d59de7d0c25a7b5924b0c600 | 12:34.39 |
| ator, sebras: Actually, better comments/commit message: https://git.ghostscript.com/?p=user/robin/mupdf.git;a=commitdiff;h=bb1cb69d1cd2d87137b153dfd12e4c787fb545a7 | 12:38.06 |
user38 | mupdf-x11 ~/RESOURCE.PDF | 12:41.19 |
| ximage: disabling shared memory extension: BadAccess (attempt to access private resource denied) | 12:41.48 |
| not sure what that ximage message is all about (it was always there, from the beginning), but other than that, mupdf-x11 seems to work, although the window is also without decorations or any other controls. | 12:42.38 |
| mupdf-gl ~/RESOURCE.PDFBus error (core dumped) | 12:43.06 |
Robin_Watts | So the 64bit version works ok for mupdf-x11, but not for mupdf-gl ? | 12:46.14 |
ator | mupdf-x11 has no UI elements like buttons or menus, it's purely mouse and keyboard driven | 12:46.16 |
| the ximage warning is just saying that it couldn't use the XShm extension to speed up copying images to screen | 12:46.46 |
| mupdf-gl crashing -- probably a bug in the opengl driver or related? a stack trace would help. | 12:47.16 |
user38 | That is correct: 64-bit mupdf-x11 works now, 64-bit mupdf-gl still crashes (it never worked to begin with on the sparc platform). | 12:47.21 |
| Let me see what I can glean out of the core file... | 12:47.56 |
ator | you might need to use the mesa software driver for mupdf-gl; I would not be surprised if the Sun's opengl implementation has bitrotted to the point of not working | 12:48.19 |
| it never was very good, even back in the late 90's early 00's when I last had to use it | 12:48.51 |
clam | ator: what wasn't? mesa's dri on sparc? | 12:49.59 |
user38 | As far as I am aware, there is no MESA OpenGL implementation on the sparc platform. | 12:50.05 |
ator | clam: sun/solaris own proprietary opengl implementation. it always threw hissyfits when you tried to make it do textures... | 12:50.33 |
| user38: I thought (but could very well be mistaken) that mesa had a pure software pipeline that worked on any X11 display | 12:51.04 |
clam | ator: hmm okay, i only had experience with sparcs x11, some of my x11 code had a bug that only manifested itself there, took blueswirl adding sparc support to qemu for me to debug it, the quick time (an hour?) i had with ultraspark in UIB in .no wasn't enough | 12:52.35 |
| ator: sure | 12:52.50 |
| ator: good luck bulding it though | 12:52.57 |
| *building | 12:53.32 |
ator | Robin_Watts: pad_size_for_pointer only pads to multiple of 2 (not 8 as I would've expected) | 12:54.14 |
clam | anywho i think mupdf-gl on soaris/sparc just wouldn't work or will require gargantuan effort... just a hunch though | 12:54.16 |
Robin_Watts | ator: size is in uint32_t's. | 12:54.31 |
ator | clam: yeah, I'm not going to waste time of it | 12:54.34 |
| Robin_Watts: ah. | 12:54.37 |
Robin_Watts | user38: so I reckon we should concentrate on making 32bit mupdf-x11 work. | 12:55.23 |
ator | if (ptr & 4) you're going to have to explain that one to me like I'm five years old | 12:55.28 |
Robin_Watts | (or, equivalently, 32bit mutool draw) | 12:55.34 |
clam | & _4_ | 12:55.50 |
Robin_Watts | if (ptr & 4) then the address is not 64bit aligned. | 12:56.05 |
clam | joins the 5yo club | 12:56.08 |
Robin_Watts | (we know the address will be 32bit aligned, hence it'll always have ptr &3 == 0. | 12:56.25 |
| but if ptr &4 then it's 32bit, but not 64bit aligned. | 12:56.39 |
| so by adding 4 we bump it to the next 64bit alignment. | 12:56.53 |
clam | Robin_Watts: &7 would make the intention less hard to see through, no? | 12:56.58 |
ator | can we not use &7 ? but then we can't just add | 12:57.10 |
clam | 5yos FTW! | 12:57.23 |
Robin_Watts | We could, but as you say, we'd then need to do: ptr = (ptr+7)&~7; | 12:57.53 |
| We'd lose the if, but gain complexity. | 12:58.04 |
| My choice would be what I wrote originally. | 12:58.24 |
user38 | @Robin_Watts: that's a plan. Should you wish to get the mupdf-gl working in the future, Alan Coopersmith from Oracle is the contact person for all things X11 and OpenGL on Solaris. He is (or was at least) in charge of porting and maintaining X and OpenGL on that operating system. @alanc on "Twitter". | 12:58.49 |
Robin_Watts | does not use Twitter, on principle. | 12:59.08 |
user38 | At least you have a name and both of you use e-mail. | 12:59.28 |
user38 | shrugs | 12:59.46 |
Robin_Watts | Yeah, thanks. | 12:59.52 |
ator | Robin_Watts: but you use facebook? Hypocrite! :P | 12:59.52 |
Robin_Watts | ator: facebook's model is "Here is stuff I've splashed upon a wall for my friends to read." | 13:00.15 |
clam | doesn't use twitter because that means turning on js | 13:00.33 |
Robin_Watts | Twitter's model is "Every asshole gets a megaphone and a roof to stand on." | 13:00.38 |
| but anyway... | 13:01.17 |
Robin_Watts | is finally allowed to get lunch. bbs. | 13:02.04 |
clam | Robin_Watts: i've watched the SP episode where Cartman is forced to quit twitter few hours ago... | 13:02.31 |
Robin_Watts | user38: If you can get a backtrace from "mutool draw -o out.png whatever.pdf" where it crashes, that'd be great. thanks. | 13:02.45 |
| (from a debug build) | 13:02.54 |
user38 | I believe that on 32-bit, you should be okay with aligning on 32-bit boundaries, since the (Ultra)SPARC processors are natively 32-bit, even when they support 64-bit addressing and instructions. | 13:02.58 |
| Let me see if I can get a build=debug out of it. | 13:03.29 |
| We're going 32-bit build=debug now, correct? | 13:04.38 |
Robin_Watts | please. | 13:06.05 |
user38 | Gladly. | 13:08.23 |
| I think my patches broke your debug build, and this is why I think that: gcc -fPIC -mno-app-regs -O3 -ffunction-sections -fdata-sections -Wsign-compare -pipe -g | 13:18.44 |
| ...by the by, what does "fitz" stand for? | 13:25.45 |
kens | user38 I don't think its stands for anything in particular | 13:29.57 |
| It was originally inteded (AIUI) to be a replacement graphics library for Ghostscript. | 13:30.16 |
| But it never got used for that, and Tor used it for MuPDF | 13:30.32 |
| Beyond that I'm unsure as it predates my involvement | 13:30.47 |
user38 | Such bits of history are always salient to me. | 13:33.47 |
ator | originally intended to be the new graphics library for ghostscript, based on raph levien's "libart" project | 13:34.28 |
| a son-of-ghostscript+libart project, hence the "fitz" name | 13:34.43 |
| I used it for mupdf, but the project to replace the graphics library in ghostscript fizzled out | 13:35.10 |
| and I used it for mupdf because we at the time wanted a smaller project to test it with than integrating with ghostscript | 13:35.49 |
| and the first implementation of mupdf (back from 2002-ish) used libart as the renderer | 13:36.20 |
user38 | Like puzzle pieces... fascinating. | 13:37.57 |
| AR build/release/libmupdf.a | 13:38.23 |
| that's how far it came with build=debug... seems to be much slower. My guess is because of -O3. | 13:38.58 |
malc_ | ator: the "son of ... hence" does not make sense to me, my mind refuses to make a connection | 13:39.17 |
| user38: | 13:39.39 |
| damn | 13:39.43 |
| user38: -O3 in debug?? | 13:39.51 |
user38 | I can't say I get the son of ghostscript reference, as I've no context. | 13:40.20 |
kens | Fitz as a surname denotes 'son of' | 13:40.21 |
user38 | In which language? | 13:40.31 |
kens | Like 'son' or 'sen' in Swedish/Danish, or 'Mc' in Scottish or 'O' in Irish | 13:40.43 |
user38 | I never knew that. | 13:40.59 |
kens | I'm not sure which language fitz is in,its used in Norman names in England | 13:41.09 |
| Norman = from Normandy | 13:41.16 |
| in France | 13:41.22 |
malc_ | i always suspected that tor is sikrittly a scott | 13:41.23 |
user38 | Well now it all makes sense. | 13:41.49 |
ator | malc_: old french, same word stem as "fils" (boy) | 13:41.51 |
kens | Ah Wikipedia says its Old French | 13:42.09 |
malc_ | ator: just yesterday i watched SP and there was a breton song at the end, loved the thing... | 13:42.56 |
ator | kens: Well, Normans spoke old french :) | 13:43.27 |
malc_ | and i recall vangelis having a song called 'la petite file a la mer' or something but there file means girl | 13:43.41 |
| FTS basically | 13:43.47 |
kens | Yeah which is why I mentioned it, I just wasn't sure enough to say that without checking | 13:43.48 |
ator | and consequently most of british nobility for a couple of centuries... | 13:43.50 |
kens | Nah we have Germans now | 13:43.58 |
ator | malc_: fille is girl/daughter, fils is son | 13:44.47 |
malc_ | kens: may i remid you that you are sottish ... and those germans are yours only through marriage | 13:45.04 |
| ator: tack! | 13:45.15 |
kens | Indeed they are, I never said otherwise :-) | 13:45.32 |
| Especially since I actually live in England | 13:45.43 |
ator | kens: I got the new Britannia board game from kickstarter the other day... but I doubt I'll be able to play it this year :( | 13:46.05 |
kens | Hmm, don't know that one let me look | 13:46.22 |
| But yeah,cancelled all gaming for now :-( | 13:46.46 |
| First published in 1986, wow.... old game | 13:47.07 |
ator | it's an oldie, been out of print for a while | 13:47.14 |
kens | 4 hours to complete, oof | 13:47.39 |
ator | kens: that's if you play quickly! first time I played I think we were at it for 6+ hours | 13:47.56 |
kens | Yeah I was assuming that it would really be longer | 13:48.11 |
malc_ | nerd alert | 13:48.11 |
ator | malc_: you're on IRC, that should be alert enough! | 13:48.28 |
malc_ | hah :) | 13:48.36 |
kens | :-) | 13:48.45 |
ator | kens: just got to thinking about it when talking about all the peoples who came and went through britain | 13:49.23 |
kens | Yes, I can see why from the description | 13:49.35 |
| covers a load of history | 13:49.46 |
malc_ | before celtic scandinavian roman and germanic barbars invaded it was such a lovely island... https://www.youtube.com/watch?v=2TR0gaG01do told me so | 13:54.28 |
ator | malc_: wasn't even an island if you go back a bit ... https://en.wikipedia.org/wiki/Doggerland | 13:56.36 |
malc_ | ator: we can go even further to the Pangea time while at it :) | 13:57.21 |
ator | well, 6000bc is just barely prehistory | 13:57.59 |
kens | It looks fairly Celtic | 13:58.38 |
malc_ | ator: true that | 13:58.48 |
kens | Oh, supposed to be Pictish | 13:58.55 |
malc_ | kens: yep | 13:59.00 |
ator | don't trust games (and definitely not hollywood) on historic costumes | 13:59.18 |
malc_ | the whole blood bath that scotts was immortalized in Stevensons(sp) poem | 13:59.31 |
ator | they put mel gibson in pictish blue paint for something that was supposed to happen in the 14th century | 13:59.41 |
kens | wikipedia says set in late 8th century, so not Pictish then IIRC | 13:59.42 |
malc_ | ator: they? | 14:00.07 |
| i beth those were jews ambushing poor mel | 14:00.23 |
ator | someone involved in braveheart. | 14:00.28 |
malc_ | also William Wallace marries Anna Karenina... and that does not sit well with me | 14:01.12 |
| that said, i loved the movie when it first came out | 14:01.25 |
| s;married;had an illicit affair | 14:01.58 |
| ator: btw have you seen my artifex github mirror submodule comment yesterday? | 14:05.17 |
Robin_Watts | user38: -O3 for a debug build is bad. | 14:05.43 |
ator | malc_: no | 14:12.28 |
user38 | Yes, it's suboptimal for a debug build. | 14:13.10 |
malc_ | ator: gist: recursive clone form github bails | 14:13.11 |
| from ghostscript - works | 14:13.36 |
user38 | Unless the source code is designed to emit extra information to stderr or stdout, I hadn't looked at it that closely? | 14:13.42 |
ator | malc_: Robin_Watts: chrisl: ah, probably those tesseract submodules that haven't been taken up by the artifexsoftware github clone | 14:14.14 |
malc_ | ator: yep | 14:14.28 |
Robin_Watts | ator: yeah. | 14:14.28 |
user38 | LINK build/release/mupdf-gl | 14:24.25 |
| ld: fatal: library -lX: not found | 14:24.35 |
| ld: fatal: library -lXrandr: not found | 14:24.43 |
| ld: fatal: file processing errors. No output written to build/release/mupdf-gl | 14:24.58 |
| so, it looks like build-debug is a no-go. | 14:25.25 |
| s/-debug/=debug/ | 14:25.36 |
| but if you can confirm to me that when make build=debug does not turn on extra telemetry output in the program, I can tell my build system to build for debugging, which will build with -O0 -g -g3 et cetera. | 14:26.49 |
| s/when make/make/ | 14:27.45 |
Robin_Watts | user38: WHy not just build mutool? | 14:28.35 |
| That avoids all the X bullshit. | 14:28.48 |
user38 | Because I don't know how. | 14:28.50 |
| Which target should I supply to make? | 14:29.20 |
Robin_Watts | I'll check. just a mo. | 14:29.29 |
| make build/debug/mutool build=debug | 14:30.20 |
user38 | That's still compiling with -O3. | 14:31.39 |
| Probably due to the way I modified the build system. | 14:31.53 |
Robin_Watts | I have a zoom meeting now :( | 14:32.09 |
user38 | No worries. I'll let it build and come back to it later as well. | 14:32.54 |
ator | user38: or just "make tools" | 14:34.05 |
| user38: the compiler flags for each build configuration are set in Makerules | 14:35.10 |
user38 | ld: fatal: file processing errors. No output written to build/release/mupdf-gl | 15:42.27 |
| It still still tried to build a release version of everything. | 15:42.44 |
| But that's okay, I don't want that distracting us. I'll tell my build system to build a debug version of mutool. | 15:43.37 |
ator | "make build=release tools" should *not* be trying to build mupdf-gl | 15:44.04 |
user38 | I did not run that yet. | 15:45.34 |
malc_ | Thread 6 "llpp.bin" received signal SIGSEGV, Segmentation fault. | 15:47.09 |
| [Switching to Thread 0x7fffed50e640 (LWP 144731)] | 15:47.09 |
| 0x0000555555651968 in fz_walk_path () | 15:47.09 |
| (gdb) | 15:47.09 |
| | 15:47.12 |
| and that's on vanilla amd64 | 15:47.21 |
user38 | Okay, so I changed build=release and the target to tools; but I also told my build system to build for debugging and thus, | 15:48.44 |
| gcc -fPIC -mno-app-regs -O0 -feliminate-dwarf2-dups -grecord-gcc-switches -DDEBUG -g -g3 -ggdb -fvar-tracking -fvar-tracking-assignments | 15:48.50 |
| ld: fatal: file processing errors. No output written to build/release/mupdf-gl | 16:04.10 |
| ...interesting. | 16:04.15 |
| That should not have broken at all. Must have missed "a spot" when forward-porting my patches. | 16:04.38 |
| For some reason, the MuPDF build engine is ignoring X11_LIBS being passed in. I'll have to investigate this further. That is the cause of the linking errors. | 16:09.24 |
| Ah, but the tools target has been built after all. | 16:12.10 |
| so running mutool draw -o out.png document.pdf produces: | 16:18.13 |
| status: process terminated by SIGBUS (Bus Error) | 16:18.21 |
| > $ccmsSetAdaptationState+0x34(254e6d0, bff00000, 0, 2550150, 401a, 1) | 16:18.41 |
| cmsSetAdaptationState+0x34(254e6d0, bff00000, 0, 2550150, 401a, 1) | 16:18.56 |
| is at the top of the stack in the core dump. | 16:19.06 |
| And also as before, if I run mutool on the one page PDF which contains no graphics, it generates out.png. | 16:21.24 |
malc_ | ator, Robin_Watts, sebras: https://tpaste.us/jnDK | 16:39.32 |
sebras | malc_: that doesn't look good, I'm assuming this is fallout of Robin's patch..? | 16:40.22 |
malc_ | user38: i can reproduce and debug path thing (probably related to yours) on a vanilla amd64 | 16:40.23 |
| sebras: guess so yeah | 16:40.39 |
| https://www.boblycat.org/~malc/scratch/text.html is the source file | 16:44.52 |
| nuc:/tmp | 16:45.00 |
| - ~/x/rcs/git/mupdf/build/debug/mupdf-x11 ~/.cache/llpp/fonts/text.html | 16:45.00 |
| mupdf-x11: source/fitz/path.c:152: fz_packed_path_size: Assertion `"This never happens" == NULL' failed. | 16:45.00 |
| zsh: abort (core dumped) ~/x/rcs/git/mupdf/build/debug/mupdf-x11 ~/.cache/llpp/fonts/text.html | 16:45.00 |
| | 16:45.03 |
| so llpp is just a bystander | 16:45.08 |
Robin_Watts | malc_: That's probably something getting out of step between writing and reading of the display list w.r.t. padding. | 16:51.06 |
| I'll look into it. | 16:51.12 |
malc_ | confirmed, resetting mupdfs git to the version just prior Robins commit fixes things | 16:51.15 |
| Robin_Watts: fab | 16:51.29 |
user38 | Is the vanilla amd64 running a GNU/Linux OS variant? | 16:51.39 |
malc_ | user38: yes | 16:51.48 |
user38 | I am surprised by this. | 16:52.01 |
Robin_Watts | user38: I need a debug backtrace with line numbers etc as a minimum. | 16:52.05 |
malc_ | user38: and please prefix your messages with the nick of the person whom you are addressing, this makes notifications possible | 16:52.24 |
user38 | malc_: like so? | 16:52.47 |
malc_ | user38: yes, thank you | 16:52.56 |
user38 | Do regular IRC II /COMMANDS work with this web IRC client | 16:53.35 |
| ? | 16:53.38 |
| Oh! They do! | 16:53.59 |
| Robin_Watts: source code line numbers? | 16:55.15 |
| Robin_Watts: 0x003289ec in cmsSetAdaptationState (ContextID=0x254e6d0, d=-1) at thirdparty/lcms2/src/cmsxform.c:67 | 16:57.07 |
Robin_Watts | user38: Right, so is the backtrace somewhere? (Sorry, I'm being pulled about 8 different ways at the moment) | 17:01.07 |
| user38: It's in the bug. gottit. ta. | 17:01.33 |
user38 | RobinWatts, is that stack trace what you want though? | 17:02.12 |
| Robin_Watts, is that stack trace what you want though? | 17:02.41 |
Robin_Watts | If that's where it's SEGVing, yes. | 17:03.33 |
| though I am confused. | 17:03.38 |
| FOr me, line 67 is: if (d >= 0.0) { | 17:03.56 |
| which doesn't involve a memory access. | 17:04.10 |
user38 | For me as well. | 17:04.11 |
Robin_Watts | What is the value of ptr ? | 17:04.32 |
user38 | $1 = (_cmsAdaptationStateChunkType *) 0x254f16c | 17:05.27 |
| The IRC client turned * ) into a smiley... | 17:06.04 |
| $1 = (_cmsAdaptationStateChunkType * ) 0x254f16c | 17:06.28 |
| Robin_Watts, is there something else I could provide before I check out for the night? | 17:09.43 |
chrisl | Robin_Watts: You might need to also be setting CMS_PTR_ALIGNMENT for the benefit of lcms2mt | 17:10.14 |
| It defaults to "sizeof(void *)" | 17:10.28 |
Robin_Watts | chrisl: Is that not enough? | 17:10.43 |
| What do we set it to for gs? | 17:10.48 |
chrisl | It's not enough on SPARC, no | 17:11.10 |
Robin_Watts | user38: OK, so that's supposed to be a pointer to a 64bit float, so clearly that's badly aligned. | 17:11.16 |
chrisl | With gs, building for sparc, we force CMS_PTR_ALIGNMENT to 8 | 17:11.41 |
Robin_Watts | user38: So you could try building with CMS_PTR_ALIGNMENT 8 tomorrow :) | 17:12.06 |
| chrisl: Thanks. | 17:12.21 |
chrisl | NP. I'll crank my machine up in the morning, and have a poke at it. No way I can face it just now.... | 17:13.00 |
malc_ | Robin_Watts: do you not see segfault on windows? | 17:13.03 |
chrisl | (If I remember) | 17:13.15 |
Robin_Watts | malc_: I have about 8 different people pulling me to do different things at the moment. I haven't yet had a chance to look. | 17:13.30 |
malc_ | Robin_Watts: sorry, nice number though | 17:14.14 |
user38 | Robin_Watts: I shall do that, then check in here tomorrow morning my time. | 17:15.30 |
Robin_Watts | user38: Thanks. What timezone are you in? | 17:15.45 |
user38 | Robin_Watts: CET. You? | 17:16.07 |
Robin_Watts | UK, so close. | 17:16.13 |
user38 | Robin_Watts: I'm UTC+2 right now. Been at it non-stop since 06:38 this morning, so I'm a little worn out. | 17:17.17 |
Robin_Watts | I understand. | 17:18.33 |
user38 | Thank you for all your support so far, and I wish you all a pleasant evening (or the rest of the day, which time zone you are in). | 17:20.01 |
| <<<Back 1 day (to 2020/10/19) | Forward 1 day (to 2020/10/21)>>> | |