| <<<Back 1 day (to 2016/11/03) | 20161104 |
hamsheet | ray_pc: thanks man I will try those | 02:21.44 |
| is it possible to use gs to convert a pdf to djvu? | 02:22.33 |
kens | hamsheet, Ghostscript does not produce DejaVu files. Its certainly possible to write a device to do so, how much time have you got to spend on the project ? | 08:02.56 |
chrisl | There's already a djvu device for Ghostscript ( http://djvu.sourceforge.net/gsdjvu.html ) but we can't include it in our distro due to license incompatibility (plus the djvu patent thing, of course) | 08:06.37 |
kens | I'd forgotten that, because its not part of the standard package... | 08:07.05 |
sebras | tor8: so array indexes should always be size_t, right? | 13:37.01 |
| tor8: why? | 13:37.03 |
| if that is the case I think eg. the array indexing in gif_read_ae() is wrong. | 13:37.52 |
Robin_Watts | sebras: Array indexes can be ints, or unsigned ints, or size_t's etc. C is not fussy. | 13:37.57 |
sebras | Robin_Watts: yes, but you guys discussed something concerning size_t and indexing..? | 13:38.18 |
Robin_Watts | sebras: For streams. | 13:38.40 |
| Streams have a position within them. | 13:38.57 |
| For 32bit builds, you'd expect that to be a 32bit unsigned number, and for 64bits, a 64bit unsigned number, but that doesn't work out. | 13:39.27 |
| cos the position can correspond to positions within a file, and for LARGEFILE builds, we can have 32 bit builds that have 64bit file positions. | 13:40.01 |
| Hence fz_off_t, which is the appropriate sized thing. | 13:40.19 |
| Hence a bunch of casting between size_t and fz_off_t etc, which tor8 doesn't like. | 13:40.34 |
sebras | Robin_Watts: ok, so size_t was introduced to resolve these issues on windows, since the issues are not visible in linux..? | 13:43.05 |
| 64-bit windows. | 13:43.17 |
| Robin_Watts: so the fix you did in load-pnm.c concerning the UINT_MAX. maybe that could be an INT_MAX instead..? | 13:49.27 |
| Robin_Watts: all types are actually int, so it would make more sense..? | 13:49.56 |
| tor8: why is the standard type in load-tiff.c unsigned and not int? | 13:51.48 |
| tor8: it seems not to match the rest of the code. | 13:52.00 |
tor8 | sebras: the spec uses unsigned ints, almost never signed. | 13:56.09 |
sebras | tor8: ok. | 13:56.33 |
tor8 | do not ask me why profilesize, tilestride or stride are signed though... | 13:56.38 |
sebras | needs to leave. | 14:02.46 |
Robin_Watts | sebras: size_t is the correct type to use for certain things. | 14:22.55 |
| pixmaps should have w/h as unsigned int's really, I reckon. | 14:23.15 |
| but god knows what knock on effects that will have. | 14:23.48 |
| tor8, sebras: Which, if any of my commits are good to go, please? | 15:19.02 |
| I'd *really* like to get the jpeg one in. | 15:19.15 |
tor8 | Robin_Watts: I may gripe about the typecasts, but all look good to me. | 15:21.32 |
| Robin_Watts: have you looked over fred's 'fix iOS app issue where deleted annotations don't properly disappear'? | 15:21.50 |
Robin_Watts | tor8: I have not. | 15:22.02 |
tor8 | that should probably be cherry-picked into the release as well, if it looks reasonable | 15:22.07 |
| my recollection of how the iOS viewer works is a bit too fuzzy to helpfully review that commit | 15:22.22 |
Robin_Watts | looks | 15:22.34 |
tor8 | do those changes stop it using the 'dirty' list of rectangles completely? | 15:23.29 |
Robin_Watts | It looks to me like it's always just plotted the whole screen. (rect is always the whole page) | 15:25.25 |
| but now it doesn't call updatePixmap any more. The diff is confusing. | 15:25.58 |
tor8 | if we're not using the dirty rect/updated rect stuff, we should remove the unused code | 15:26.06 |
Robin_Watts | mmm. Seems a shame to always be working in whole pages. | 15:26.24 |
tor8 | this commit looks a bit like a stop-gap measure, but I can't say I fully understand it | 15:26.28 |
Robin_Watts | I agree. | 15:26.35 |
| and feel the same. | 15:26.40 |
| Forward 1 day (to 2016/11/05)>>> | |