| <<<Back 1 day (to 2020/12/29) | Fwd 1 day (to 2020/12/31)>>> | 20201230 |
afxgroup | evening guys, i'm trying to port MuPDF on AmigaOS4 using glut at moment | 18:26.56 |
| the program starts correctly but when i try to open a file i get a DSI in some operations that using 64bits | 18:27.30 |
| First of all. Is Big Endian supported? | 18:27.53 |
| because AmigaOS on PPC and 68k is big endian | 18:28.23 |
| i didn't find nothing where i can set if the platform is big endian or not. Is MuPDF endian agnostic? | 18:28.44 |
sebras | afxgroup: do you have a backtrace of where this happens? | 18:32.48 |
| I haven't heard that we've had endianness issues in a long time. | 18:33.16 |
afxgroup | well. i have the exact line where it crash | 18:33.18 |
| and is | 18:33.22 |
| if (len > (int64_t)((INT64_MAX - t) / n)) | 18:33.23 |
| in pdf-xref.c | 18:33.31 |
| my len value is "20" | 18:33.50 |
| sorry | 18:34.02 |
| my "n" value | 18:34.05 |
sebras | what about len and t? | 18:34.20 |
afxgroup | len = 1776 | 18:34.44 |
| t = 701378 | 18:34.53 |
sebras | right so you should get 1776 > LARGEVALUE and that ought to evaluate to false. | 18:35.32 |
| btw, what is a DSI? | 18:36.27 |
| data storage interrupt? | 18:36.58 |
afxgroup | https://www.papersoft.it/t.jpg | 18:37.06 |
| DSI is like a null pointer exception | 18:37.13 |
| A DSI exception occurs when a data memory access cannot be performed for | 18:37.38 |
| yes is data storage interrupt | 18:38.07 |
| i'm trying to debug the values and i have really strange values when i read files. For example in your fz_atof function you call in the stack trace the divide function | 18:39.36 |
| and i get an error there. If I use strtof instead your fz_atof function i have no more errors | 18:40.05 |
sebras | len, t and n are all stackallocated, so alignment issues shouldn't be an issue. | 18:40.06 |
| though I don't know if mupdf has ever been used with newlibc. | 18:40.24 |
afxgroup | that's why i suspect there is something in the endian that is causing the problem | 18:40.27 |
| i could try to swap my libc | 18:40.41 |
| just to see if the problem is the same but before this. Is there a switch or a define to use to enable/disable big endian or little endian? | 18:41.16 |
sebras | but regardless of what values len, t and n has there shouldn't be a problem? | 18:41.20 |
| maybe if n == 0, because of division by zero. | 18:41.39 |
| but that's not the case here. | 18:41.44 |
afxgroup | i don't know but i'm receiving strange values. maybe my values are trashed. | 18:41.59 |
sebras | afxgroup: n == 20 seems reasonable. | 18:42.16 |
afxgroup | yes | 18:42.36 |
sebras | I'd expect t to be about the same size as your file. | 18:42.41 |
| afxgroup: are you able to share the file? | 18:44.16 |
afxgroup | the pdf file? | 18:44.25 |
sebras | yes | 18:46.48 |
afxgroup | https://www.papersoft.it/AISS47.pdf | 18:47.16 |
sebras | on my system I get len == 1776, t == 701378 and n == 20 | 18:49.05 |
afxgroup | so is the same | 18:49.21 |
sebras | but I can successfully open the file, so the input values are correct. | 18:49.34 |
afxgroup | ok. so is not an endianess problem | 18:49.47 |
| i'll try to use a different libc instead of newlib | 18:50.01 |
| just to see if something change | 18:50.10 |
sebras | afxgroup: can you try one thing for me? can you do these changes to the source code and recompile? https://pastebin.com/raw/BU0XcaRc | 18:52.50 |
afxgroup | yes | 18:53.42 |
sebras | afxgroup: I don't understand wht the issue is yet, but I'm thinking since the comparison is between int and int64_t maybe that causes an issue. | 18:55.17 |
| the values of the variables oughts to be the same though! | 18:55.42 |
afxgroup | indeed | 18:59.13 |
| is the same | 18:59.14 |
| let me force the align in the compiler | 18:59.31 |
| no way. well. thank you for your time | 19:02.05 |
| i'll try to find what is wrong | 19:02.12 |
| i'll try to use a different clib and see if something change | 19:02.23 |
| happy new year :) | 19:04.19 |
| well | 19:21.26 |
| https://www.papersoft.it/t1.jpg | 19:21.27 |
| https://www.papersoft.it/t2.jpg | 19:21.30 |
| with clib2 it is working | 19:21.35 |
| however the title has trash characters | 19:21.47 |
| and if i use page down i get that assertion | 19:21.56 |
| just to inform you :) | 19:22.02 |
| i have to debug everything | 19:22.08 |
superturbocat | Hello, please tell me. Is there normal mujs documentation? And the examples on the site are outdated. | 20:41.10 |
| Well, please help me how to make a function in C that returns an object. | 21:04.20 |
| <<<Back 1 day (to 2020/12/29) | Forward 1 day (to 2020/12/31)>>> | |