[CinePaint] strange handling of SIGTERM
Gene Selkov <[email protected]> Tue, 18 Aug 2009 00:11:21 +0100
| Newsgroups | gmane.comp.video.cinepaint.user |
|---|---|
| Message-ID | <[email protected]> |
Greetings to all Cinepainters out there, I succeeded in building Cinepaint on my Ubuntu laptop for the first time, after a few failed assaults spread over a couple years. This time, I had no problems at all -- just ran ubuntu.sh and got it running a few minutes later. Instant gratification followed, lasting a couple days or so. I think I am even beginning to to understand color management and other fabulous things. I've got a few small issues, and I am not sure whether they have something to do with the way I built it (totally mindlessly), or they are features. 1. It doesn't want to die. It responds to sigterm by looping and saying "sigpipe caught". It seems as though the "script-fu" bit dies first and makes the parent process wait on a broken pipe. And the parent seems to be detached from the shell it was started from, so it doesn't any further signals from the shell. I have so far found only two ways of killing it: from the parent shell by its job number (it ignores SIGTERM, but it can be stopped and then killed). Another way to kill it is by deleting its main window through the GUI (gnome). A somewhat messy consequence of this behavior as the when my Xorg dies (and it does that often, owing to a perpetually buggy nvidia driver), or when I log out, cinepaint remains looping in this half-dead state, wasting 100% CPU. I just got two of them do the same thing, having overlooked the first instance. 2. Image windows ignore button4, button5, and whatever other scrolling and zooming events are typically used in graphics apps. 3. While zooming in or out with '+' or '-', the window does two strange things: (1) it grows; and (2), it re-positions itself. I end up chasing the window all over my screen. I wonder whether this behavior is optional and can be turned off. 4. I read that I can change menu keyboard actions by entering the desired key combination while a menu item is selected. The labels in the menu do change, but the new key codes are not recognized. And they don't stick; the next time I run it, they are gone. I'd love to be able to replace Alt wih Ctrl in most menus, because my short fingers don't stagger too well. Also, should I succeed in building Cinepaint on my mac, it has only one Alt key. 5. Many menus don't have any key codes linked to them. Is that because they never had, or did something fall through the cracks while I built it? 6. I tried to see whether tools respond to keystrokes. The text tool responded to T all right. Nothing else that I am used to in Gimp worked; I got this message in response to everything I tried, except for selection-related commands, which work fine: Error: Unable to set state for menu which doesn't exist: <Image1>/View/Toggle Rulers 7. I feel I ought to put this question in a separate post, because it is so different from everything I've observed so far, and it's by far the most important one. It may be that I haven't really begun understanding color management. Whatever I tried to do during the last couple days left me with a suspicion that my display profile is applied twice. First when it is loaded into my display's LUT. And the second time, when Cinepaint opens my image. The reason I believe so is because the same image looks very different when I view it with a non-color-managed tool. To be exact, it looks right in evince, for example, which is not color-managed. To make sure I knew what's right and what's not, I took a shot of a reflective IT8 in direct sunlight, saved it into an sRGB tiff with ufraw, using my camera's internal matrix (which is a more trustworthy calibration source than any profile I managed to produce so far), and when I view that in any profile-agnostic tool on my newly calibrated screen, it does look right. Which means, I believe, that Argyll did a great job, and that my camera isn't too bad either. Now, when I open that image in Cinepaint (regardless of the profile assigned), It looks too bright and too red -- which is what I would expect from my current display profile, if it were applied twice (I know the direction and approximately the amount of change it effects, from the uncalibrated state, and the change I observe in Cinepaint is in the same direction, and by about as much). Another curious observation is that when I open my raw file with Cinepaint, it does look right, but only until I try to something useful to it. I cannot save it or make any changes, or add a layer, or copy it to another image; whatever I do to it make it look wrong. It becomes too dim and too blue. Finally, the most bizarre thing I observed: I pretended to ignore all the circumstances leading to this apparent double-correction and did some curve editing in Cinepaint to force image back to where I would like it to be, and I did that in Lstar-RGB (although the difference would be only slight if I did it in sRGB; then I saved it and opened it in Krita and a number of other color-managed tools. It looked the same in all of them. It even looked the same in Gimp, after conversion to sRGB that Gimp offers when it sees an unknown profile. Without such conversion, the image looked too blue and too dim. I was encouraged by that, and decided to send the image I edited in Cinepaint to a print shop. If it looks right in all those color-managed viewers, why wouldn't it print well? The shop's proprietor got back to me saying that the image is uncomfortably blue, and it didn't matter whether he kept Lstar-RGB or converted it to sRGB. There is something profoundly wrong in my toolchain, because when I consider all these effects together, they add up to something like Escher's straircase. Does any of this seem familiar to anyone? Is there a rational way of sorting it out? Thanks, --Gene ------------------------------------------------------------------------------ Let Crystal Reports handle the reporting - Free Crystal Reports 2008 30-Day trial. Simplify your report design, integration and deployment - and focus on what you do best, core application coding. Discover what's new with Crystal Reports now. http://p.sf.net/sfu/bobj-july