| <<<Back 1 day (to 2012/07/28) | 2012/07/29 |
mvrhel_laptop | Robin_Watts: That is an odd one. I would do your change to 32 and open a bug about the < 32 case and assign it to me | 03:23.34 |
| I am talking about number 20 in your bmpcmp | 03:24.07 |
malc_ | build/release/mupdfshow.o: In function `pdfshow_main': | 15:19.25 |
| mupdfshow.c:(.text+0x804): undefined reference to `pdf_print_xref' | 15:19.25 |
Robin_Watts | malc_: Clean and rebuild ? | 17:01.56 |
thaxy1 | hi there could someone please help me to build the commands i need | 17:12.47 |
| could please someone help me with that http://stackoverflow.com/questions/11711669/ghostscript-convert-pdf-to-image-without-saving-to-hdd ? | 18:06.23 |
sebras | Robin_Watts: hm. sometimes I think that it might have made sense to express the dependencies between files in mupdf so that they are automagically rebuilt without the need for make nuke... | 19:23.47 |
Robin_Watts | sebras: Sorry? The dependencies should be correctly expressed in the makefile. | 19:24.23 |
| If they aren't, then that is a bug. | 19:24.41 |
sebras | Robin_Watts: so how come malc_ saw that undef ref? | 19:25.26 |
Robin_Watts | editing files while making? changing between branches with git? | 19:26.26 |
sebras | the first I buy and doing so is a bad idea. | 19:26.48 |
| the latter shouldn't generate dependency errors as long as you are switching to a branch with dates later than the one you start out with. | 19:27.30 |
| hm... | 19:27.36 |
| Robin_Watts: hm.. there appears to be a race between when some files are built and when they are needed. | 20:14.08 |
| I'm seeing inconsistent builds on my machine using make -j10... | 20:14.33 |
| I reproduce as follows: git checkout $SHA1 && make -j10 nuke && make -j10 and store the exitcode, then git checkout $SHA1~1 && make -j10 nuke && make -j10 && git checkout $SHA1 && make -j10 and store the exit code | 20:54.44 |
| If both exit codes are zero, there is no problem.\ | 20:55.06 |
| If both exit codes are non-zero, then $SHA1 has a problem, but there is not inconsistency in the build process | 20:55.41 |
| If one exit code is zero and the other non-zero, there _is_ an inconsistency in the build process. | 20:56.01 |
| I've re-run this for all commits on master from HEAD to 1.0 a couple of times and seen that sometimes the exit codes do differ. | 20:56.39 |
| last time is about commits fc3dfd6 and d7d281d. I manage to build d7d281d correctly from scratch but not when starting out with a build at fc3dfd6 and adding d7d281d and then building incrementally. | 20:57.53 |
| d7d281d appears to add pdf_write_document to pdf/pdf_write.c and pdf_create_object to pdf/pdf_xref.c | 20:58.35 |
| in the linking process mupdfposter.c complains about them not being available. | 20:59.21 |
| the newly added functions that is. | 20:59.30 |
| running nm on the object-files reveal that pdf_write.o and pdf_xref.o contain the relevant symbols. | 21:00.28 |
| libfitz.a does not however, despite including pdf_write.o and pdf_xref.o (though probably these are the old versions built from fc3dfd6). | 21:01.19 |
| hm... even analyzing the timestamps using ls -lt --full-time doesn't reveal why the new .o-file is not included in libfitz.a | 21:33.59 |
tor8 | sebras: moving files around can leave old garbage object files i library archives that mess up subsequent linking | 21:43.18 |
sebras | tor8: so how come this only happens intermittently? | 21:47.47 |
| tor8: and that implies that ar cru ought to be ar cr to get rid of the problem... | 21:49.49 |
tor8 | sebras: hmm, no idea. | 21:50.04 |
sebras | tor8: also ar only stores hh:mm:ss, no parts ms/ns at all. which of course might mess up time comparisons. | 21:51.49 |
| I was able to retrieve the timestamps from libfitz.a using ar xo libfitz.a | 21:52.10 |
| tor8: ar cru -> ar cr looks good so far. I'll whip up a patch to make this change, so we don't have to tell people to make nuke for this reason at least. | 21:55.52 |
tor8 | right | 21:56.11 |
sebras | linking will take infinitesimally longer, but as long as it is consistent I'm much happier. | 21:56.20 |
tor8 | yeah. anyway, time to sleep. good night. | 21:58.32 |
sebras | nn. | 21:58.41 |
| tor8 Robin_Watts: two patches over at sebras/master, one fixing a typo in mudraw and another fixing the error I was writing about before. | 22:22.51 |
| I believe that my patches are sound, but I haven't been able to test the one for iOS, can you? | 22:23.15 |
pipitas | cryptopsy: You there? | 22:23.21 |
| Forward 1 day (to 2012/07/30)>>> | |