| | <<<Back 1 day (to 2015/07/29) | 20150730 |
| pavan | Hi Robin , | 06:12.02 |
| chrisl | rayjj: there are a few options which are only relevant to early or certain versions of Visual Studio, and so give such warnings in later/other versions. I *try* to make sure that they are all benign (i.e. simply ignored) when not relevant. | 07:42.56 |
| pedro_mac | morning folks | 08:09.23 |
| chrisl | Good morning | 08:10.12 |
| Robin | pavan: (For the logs) I'm on holiday at the moment, in a different timezone, so can't always reply. | 09:04.44 |
| tor8 | Hi Robin! | 09:07.24 |
| Robin | Morning tor8. | 09:07.35 |
| | Good holiday? | 09:07.39 |
| tor8 | Yes. And I see you haven't been up to quite as much mischief as I'd feared :) | 09:07.53 |
| Robin | Nah, I was occupied with other stuff. | 09:08.08 |
| | The only major mupdf stuff was GProof support. | 09:08.28 |
| tor8 | The gproof stuff is all news to me though, something that came up during the june meeting? | 09:08.29 |
| Robin | Yeah. | 09:08.34 |
| | it had been mentioned at previous meetings. | 09:08.43 |
| tor8 | I did a bit of rejigging of two patches, it's on tor/master | 09:09.00 |
| Robin | The idea is that we use a gs/mupdf combo to do on device color correct proofing. | 09:09.16 |
| | There is a twiki page with details. | 09:09.27 |
| tor8 | not thrilled about forking and calling exes, and shipping gs as well in the apps, but it's self contained and serves a good purpose | 09:09.42 |
| | but I worry about render mismatches between gs and mupdf | 09:10.08 |
| Robin | tor8: The second patch looks good. | 09:11.37 |
| tor8 | Robin: so when are you coming back from vacation? (and where are you, I forgot) | 09:11.48 |
| Robin | The first one (moving the read16le etc stuff)... I'd have been tempted to put it into a new file if you are concerned about space. | 09:12.08 |
| sebras | tor8: tahiti! | 09:12.13 |
| | tor8: I think. | 09:12.19 |
| Robin | Cos I think other things could benefit from it in future. | 09:12.23 |
| | Moorea now. | 09:12.28 |
| | (17km from Tahiti) | 09:12.36 |
| | Back in 2 weeks. | 09:12.44 |
| tor8 | Robin: if we want to keep the int reading functions public, I say we make the full set available and use them consistently | 09:12.44 |
| Robin | tor8: yeah, that would be my preferred choice. | 09:13.01 |
| tor8 | currently we have a lot of small functions like that sprinkled as static functions in the places where needed | 09:13.03 |
| | Robin: I'll do that then | 09:13.24 |
| Robin | The mismatches in rendering between gs and mupdf are the reason for the gproof device :) | 09:13.40 |
| tor8 | in all of [u]int{16,32,64} permutations | 09:13.51 |
| | both big and little endian | 09:14.00 |
| Robin | tor8: We optionally can call via gsapi, so no forking. | 09:14.15 |
| | tor8: yeah, full set seems sensible. | 09:14.26 |
| | The android stuff uses the gsapi methods. | 09:14.55 |
| tor8 | gsapi would be better I think; does this tie in with chrisl's directory restructuring? | 09:15.22 |
| Robin | It works with chrisl's new restructure, yes. | 09:15.41 |
| | I think I committed the gsapi changes... | 09:15.48 |
| tor8 | fork/pipe/exec would also be preferable to temp files, but that's probably more work than its worth | 09:15.54 |
| Robin | gsapi can probably do pipes, but at the time I was desperately trying to get a working version up and working before I went away. | 09:17.09 |
| tor8 | Robin: I have to admit, I haven't looked too closely at the gproof stuff other than reading the twiki page | 09:17.16 |
| Robin | Miles wants it for demo at the Chicago show. | 09:17.17 |
| chrisl | Robin: is the gprf format streamable? | 09:17.43 |
| tor8 | .gproof and .gprf file endings are a bit awkward, IMO | 09:17.46 |
| Robin | tor8: Henry and mvrhel have looked it over. It hasn't required many changes to mupdf. | 09:18.02 |
| | I tidied the get_pixmap stuff to be sane. | 09:18.23 |
| | And I added separations methods to fz_page's. | 09:18.35 |
| tor8 | I need to go over the separation stuff as well, and almost any change to fz_image are bound to be improvements :) | 09:18.42 |
| Robin | Everything else is hidden within the format handler. | 09:18.55 |
| | chrisl: Kinda. | 09:19.03 |
| chrisl | Kinda....? | 09:19.21 |
| Robin | It needs seeking when writing (for the table of contents) | 09:19.40 |
| tor8 | the separation interface looks good | 09:19.45 |
| chrisl | So, can't use pipes then....... | 09:19.51 |
| Robin | D'Oh. Yes. | 09:20.01 |
| tor8 | Robin: really? I thought the format didn't have much of a table of contents. | 09:22.29 |
| Robin | many images, many planes, all differenced and zlibbed. | 09:23.14 |
| tor8 | the ICC offset and Table offsets should be known at the time of writing the header | 09:23.20 |
| Robin | hence unpredictable sizes. | 09:23.27 |
| | The offset of the table is known. The contents of the table are not. | 09:23.51 |
| tor8 | ah, the tile table | 09:23.56 |
| Robin | On playback, we return to the file many times (whenever we redecode with a different set of seps) | 09:24.16 |
| tor8 | if we'd made the format chunked with chunk headers instead it could have been streamable | 09:24.17 |
| Robin | We can talk about a v2 when next we meet. | 09:24.45 |
| | Feel free to add notes on a v2 to the twiki page. | 09:24.57 |
| tor8 | but yeah, it makes sense to have it random access in the reader | 09:24.59 |
| chrisl | You probably wouldn't want to have it all in memory anyway | 09:25.20 |
| tor8 | the zip approach of a table of contents at the end as well as streamable chunks is a bit awkward, but does offer a lot of flexibility for both use cases | 09:25.45 |
| Robin | yes, we have the freedom to put the table at the end, I think, but for simplicity i didn't bother in the first version. | 09:26.21 |
| tor8 | you could emit the current format with the tile table at the end of the file, but then you'd still need to patch the tile offset pointer in the header | 09:26.47 |
| chrisl | These files have the potential to be quite large | 09:26.59 |
| tor8 | make the tile offset pointer a signed integer, and make it negative to count from the end of the file? | 09:27.09 |
| | then we could write without seeking | 09:27.28 |
| | anyway, that can be discussed in september | 09:27.45 |
| | we've got almost two full days of doing nothing after the meeting | 09:27.59 |
| Robin | Given we want to seek at the reader, I'm not sure streaming at the writer is a win. | 09:39.38 |
| tor8 | Robin: Fair enough. | 09:40.01 |
| Robin | bedtime. Night. | 09:47.43 |
| jogux | night Robin | 09:56.44 |
| henrys | tor8 did you get with fred? | 14:10.27 |
| tor8 | henrys: yes. | 14:53.39 |
| rayjj | chrisl_away: can you have a look at my VS2015 and PROFILE patches: http://git.ghostscript.com/?p=user/ray/ghostpdl.git;a=summary | 15:22.32 |
| | chrisl_away: I'm able to build with VS 2015 Community edition (free) and use the Performance Explorer to get a real profile (for the first time). | 15:23.39 |
| | I have to run an errand. bbiaw | 15:31.54 |
| chrisl | rayjj: on your VS2015 patches: the one for "bool", I'd prefer that to go in one place, rather than two separate source files. My preference would be to have it in base/stdio_.h - there is already some platform specific stuff in there, it's already included in both the places it's needed, and anywhere else it might be needed in the future. | 17:12.36 |
| | The "profile" one looks fine, although I might be tempted to suggest making the extra flag conditional on being VS2015 - but that's not critical. | 17:13.35 |
| rayjj | chrisl: I doubt the profile build for pcl/xps ever worked for Windows. VS profile tools were only on certain expensive versions and didn't work well (I could never get the one with 2008 to work -- it crashed the OS) | 17:37.37 |
| | chrisl: the build would succeed, it's just when the performance tool started that I got the message about needing the /PROFILE option for the LINK | 17:39.11 |
| | chrisl: so as far as I know, it's needed for all veresions of VS | 17:39.49 |
| chrisl | rayjj: okay, that's fine. I was thinking of the older VS versions that didn't support instrumented profiling at all | 17:40.28 |
| rayjj | chrisl: well, the option has been there for the gs build for quite a while | 17:41.07 |
| chrisl | rayjj: as I said, I wasn't too bothered about that commit - not something I'm going to make a fuss about | 17:42.37 |
| mvrhel_laptop | sebras: your commit for gif in mupdf did not update the visual studio solution | 17:44.09 |
| | so that is broken | 17:44.19 |
| | do you want to fix it or shall I | 17:44.24 |
| chrisl | would be *shocked* if sebras has Visual Studio - possibly unless it's for work...... | 17:45.17 |
| mvrhel_laptop | in that case, I will take care of it | 17:45.34 |
| | since I would like to be able to build.... | 17:45.43 |
| chrisl | You have *such* high standards! | 17:46.00 |
| mvrhel_laptop | :) | 17:46.06 |
| | I am very needy | 17:46.12 |
| chrisl | Well, as we all know, if it builds, it's ready to ship | 17:46.43 |
| | Right, I have me a squash match to get ready for..... bye! | 17:48.32 |
| mvrhel_laptop | Robin, so this is a two line change in one of the .proj files, I am going to push to golden if that is fine rather than wait for review... | 17:53.18 |
| milford | Ubuntu released a patch for CVE-2015-3228 today. Are any of the ghostscript downloads at http://www.ghostscript.com/download/gsdnld.html vulnerable to this security problem? If so, is there some way to get a patched binary? | 18:04.13 |
| rayjj | chrisl_away: the common place for the undef bool is actually windows_.h -- it has to follow the #include <windows.h> that #defines bool | 18:27.31 |
| | but I figured you'd be satisfied if it's in just one place | 18:27.56 |
| mvrhel_laptop | I guess I need to set up gsview and mupdf with gproof file handling to make this easier for me to test | 19:01.09 |
| | it appears that mudraw is not really set up for it directly | 19:01.17 |
| | I need to get a few unfinished things in gsview done and committed first though | 19:04.42 |
| Robin | I've been testing it with mudraw. | 19:15.27 |
| | lots of signed/unsigned warnings in load-gif.c | 23:14.39 |
| sebras | Robin: oops. | 23:31.55 |
| | Robin: let me take look. | 23:32.04 |
| | Robin: strange. I see none of those warnings using gcc. | 23:32.50 |
| | oh, so I need to compile using XCFLAGS=-Wconversion to see those. | 23:35.09 |
| | Robin: or perhaps -Wsign-conversion? | 23:53.17 |
| | Robin: you see those in visual studio, right? | 23:53.40 |
| | Robin: do you see similar warnings in load-png.c? | 23:54.05 |
| | Forward 1 day (to 2015/07/31)>>> | |
IRC Logs