| <<<Back 1 day (to 2016/04/28) | 20160429 |
kens | On the move chris_x250 ? | 08:21.33 |
chris_x250 | I've got some work being done on the house - so I was either stuck *in* my office, or stuck *out* of the office - I chose out..... | 08:22.21 |
kens | Seems reasonable | 08:22.32 |
| Can I bother you for a sanity check ? | 08:22.38 |
chris_x250 | Sure | 08:22.58 |
kens | assuming you have access to the GS source ? | 08:23.01 |
chris_x250 | I do | 08:23.06 |
kens | OK in base/gxshade.c, line 374 | 08:23.22 |
| for (ci = 0; ci < pfs->num_components; ++ci) | 08:23.44 |
| On the first pass through that loop, what is the value of ci ? 0 or 1 ? | 08:23.57 |
chris_x250 | 0 | 08:24.10 |
kens | That's what I thought, then I confused myself and couldn't figure it out, thanks | 08:24.28 |
chris_x250 | I'd be tempted to change that to the more conventional postfix increment | 08:24.54 |
kens | Well it is slightly confusing that way (says the confused bear) | 08:25.18 |
chris_x250 | Exactly, yes. | 08:25.38 |
kens | In any event, the Coverity warning is a false positive, so I'm happy, thanks | 08:25.56 |
chris_x250 | A bit concerning that coverity would be confused by that...... | 08:27.06 |
kens | sorry, network hiccup | 08:28.04 |
chris_x250 | I just sais: A bit concerning that coverity would be confused by that...... | 08:28.23 |
| or even said | 08:28.27 |
kens | The Coverity problem is that 'ranges' is set based on the type of a CIE space, and can be arrays 3, 4 or a single pair depending | 08:28.34 |
| It sets tanges to the address of a pair, then uses that as if it were an array in that loop. So if num_components wasn't 1, it would overrun | 08:29.11 |
| THat's what it warnign about | 08:29.18 |
| Of course we cannot get here with num_components != 1 if its a CIEBasedA space | 08:29.33 |
| So its a false positive | 08:29.44 |
chris_x250 | Right - I suspect there will be a lot of those..... | 08:30.07 |
kens | Surprisingly few so far, though I have skipped a number that may well be like ths. The confusion was entirely mine when I wondered if ci could be > 0 and tehn tied myself in knots trying to figure it out | 08:30.44 |
| And we don't have a file in the test suite whch exercises ths path apparently | 08:31.13 |
| I guess CIEBasedA shadings are kind of unusual :) | 08:31.38 |
chris_x250 | That is another problem - I've found a fair few like that.... | 08:31.43 |
kens | Henry has expressed concern about it, it would be possible to write test cases I guess. I certainly could for ths particular one | 08:32.41 |
chris_x250 | There are some in the font code we definitely can't write cluster tests for..... | 08:33.20 |
kens | I guess that's less surprising | 08:33.37 |
chris_x250 | Well, for consistent results across machines we cannot allow gs to do "clever" font substitutions | 08:34.24 |
kens | Yeah I know, sad but true. | 08:34.47 |
chris_x250 | I guess we could probably do it for weekly tests: I think those always run on a single machine | 08:35.34 |
kens | I'm not sure its really worth it. There are so many options in GS we can't test them all anyway | 08:36.12 |
chris_x250 | Actually, for testing the font substitution stuff, we probably just need to run the normal cluster tests without disabling the native font usage. | 08:37.10 |
kens | I guess we have enough PDF files with missing fonts ? :-) | 08:37.34 |
chris_x250 | Oh god yes! | 08:37.46 |
kens | Ah I thnk Coverity has found a real bug here..... | 08:37.54 |
| Actually I take back that comment about test files. I was looking at the wrong regression report (I have so many....) Turns out it *is* tested, 4 files from the CET | 08:39.52 |
chris_x250 | Well, I cannot work out why authentication with freenode isn't working on this laptop - so I'm not going to be able to see our other channel for a while | 08:52.02 |
kens | chrisl yeah that's why I pinged you here when you came on, I'd already done so on the other side | 08:57.45 |
sebras | tor8: I fix the makefile for cmapdump over at sebras/master | 09:34.10 |
tor8 | sebras: LGTHM. | 09:42.23 |
chris_x250 | I'm guessing we don't use macgenmcpxml.sh any more..... | 09:44.37 |
kens | Never heard of it | 09:51.25 |
sebras | tor8: to humble me? | 09:56.09 |
Robin_Watts | LGTHM? Is that like LGBTQCI ? | 10:42.50 |
chris_x250 | I might start using: LGTMBTBQHIHGAFC | 10:45.23 |
Robin_Watts | My mail server is down. Apaprently ix had a series of drives in the RAID unit fail :( | 11:00.07 |
kens | All you're missing at the moment is my dull series of Coverity commits | 11:02.28 |
sebras | chris_x250: Looks Good To Me But The Boneheads Question Html in Hashtables Going Against Function Calls? | 11:03.04 |
chris_x250 | sebras: "Looks Good To Me But, To Be Quite Honest, I Haven't Got a F***ing Clue" | 11:03.32 |
sebras | chris_x250: oh... | 11:04.04 |
chris_x250 | Specifically to be used when asked to review some changes in area you've never even looked at before...... | 11:05.04 |
kens | I think I need that acronym | 11:05.29 |
| But its too long, I'd never remember it all | 11:05.49 |
chris_x250 | Possibly can be shortened to: "yeh, whatever......." | 11:06.14 |
sebras | kens: btw, http://bugs.ghostscript.com/show_bug.cgi?id=696687 is ready for gs now. | 11:06.17 |
kens | I'll assign it to me to look at then | 11:06.36 |
| Probably you should put in a comment saying its fixed for MuPDF first though | 11:06.54 |
Robin_Watts | chris_x250: I've been using "looks plausible" for that for years. | 11:07.16 |
kens | LP ? I two letter acronym I can probably live with | 11:07.40 |
sebras | kens: done. | 11:08.39 |
kens | RIght then I'll assign it to me, it'll be something to do this afternoon | 11:08.56 |
sebras | kens: I hope the fix for gs is as easy as for mupdf | 11:09.58 |
kens | I have no idea yet I'm hopeful though. Although nothing is ever easy when you are talking about Ghostscript, and in particular its PDF interpreter........ | 11:10.41 |
| Drat a stupid change that was so small I didn't test it has caused problems, of course..... | 12:14.18 |
| Forward 1 day (to 2016/04/30)>>> | |