| | <<<Back 1 day (to 2017/11/28) | 20171129 |
| deekej | hello guys, am I correct that you do not use the Makefile.ac, just the Makefile.in for ghostscript? | 13:39.31 |
| | (IOW, your Makefile.in files are written by hand?) | 13:39.51 |
| chrisl_laptop | Yes | 13:40.43 |
| deekej | thanks | 13:40.47 |
| chrisl_laptop | Hence we use autoconf, and not automake | 13:41.38 |
| deekej | I'm trying to extend the Makefile.in & configure.ac with new parameter, something like --without-versioned-path | 13:42.43 |
| | it should ease up the packaging at least for Fedora, maybe other distros as well | 13:43.16 |
| chrisl_laptop | Um, that probably isn't a good idea | 13:43.18 |
| deekej | chrisl_laptop: are there any places where it could backfire? | 13:43.34 |
| | currently I tried to do this the dirty way, while updating the ghostscript's search path for the new location | 13:44.18 |
| chrisl_laptop | People regularly have more than one verison of Ghostscript installed. And the init files and not always compatible with other veresions. | 13:44.23 |
| deekej | chrisl_laptop: yes, I realize that. :) This might not be beneficial for other distros than Fedora. | 13:45.26 |
| chrisl_laptop | I'm not sure I see how it's beneficial for Fedora..... | 13:46.03 |
| deekej | In Fedora, there's policy that only one package can be installed at a given time. If 2 or more different versions are needed to be installed at the same time, then they should be containerized. | 13:46.07 |
| | and for Fedora, there's a requirement that at least the documents for each package should be located in non-versioned path | 13:46.48 |
| chrisl_laptop | Most distros just symlink the versioned path to "current" or similar | 13:48.18 |
| deekej | on Fedora, users are expected to use the packages officially provided. If they do not, they are doing it at their own risk. IMHO, one of the ideas behind this is that users should report any problems to BZ, so they are also resolved for others. | 13:48.25 |
| chrisl_laptop | Trouble is, the version numbers in the Makefile don't come from configure, they come from another makefile | 13:49.54 |
| deekej | chrisl_laptop: you mean the base/version.mak? | 13:50.22 |
| chrisl_laptop | Yes | 13:50.29 |
| deekej | ok, but theoretically, the ghostscript should work okay if only one version of it is installed at the time, right? | 13:51.16 |
| chrisl_laptop | Yeh | 13:51.51 |
| deekej | chrisl_laptop: ok, I will try to create some elegant solution to this, which would also contain a warning that using it is user's own responsibility & risk | 13:52.53 |
| chrisl_laptop | So, in Makefile.in we have assignments for gsdatadi and gssharedir...... | 13:53.12 |
| deekej | I'll post it to you after, and you can decide if you would include in in ghostscript or not :) | 13:53.17 |
| | I was thinking about creating the option I mentioned inside the configure.ac, and putting Makefile's if-statement into the Makefile.in, to the line you are mentioning. | 13:54.12 |
| chrisl_laptop | No, that'll break every make except gnu make | 13:54.43 |
| deekej | ah | 13:54.52 |
| chrisl_laptop | Those finish with $(GS_DOT_VERSION) - you could have configure set that final part of the path to either be "$(GS_DOT_VERSION) " or something else. The escaping will be a pain, but... | 13:55.16 |
| deekej | my dirty solution so far was "emptying" the $(GS_DOT_VERSION). It seems it worked, however creating something like this: | 13:56.47 |
| | search path: /usr/share/ghostscript//Resource/Init | 13:57.00 |
| chrisl_laptop | Leaving GS_DOT_VERSION unset would be, imho, unwise - we could well decide to use it for something else at some point | 13:58.58 |
| deekej | I see. I also don't like that solution. I used it just to test my initial thought. | 13:59.39 |
| | I'm checking some portable ways for conditionals for both GNU make and BSD make | 14:00.12 |
| chrisl_laptop | Okay, and Sun make? And HP make? And SGI make? etc, etc | 14:02.20 |
| deekej | :-/ | 14:02.37 |
| | I actually didn't know Sun & HP had their own make... o.O | 14:02.57 |
| chrisl_laptop | That's one of the reasons for some of the crazier stuff in our sub-makefiles | 14:05.23 |
| deekej | chrisl_laptop: I see. Hmm, I'll see if I can come up with something, I'm not giving up without a fight... :D If not, I'll just keep some well documented downstream patch for Fedora, and remove it if something doesn't work. | 14:06.52 |
| chrisl_laptop | deekej: give me a half hour or so, I'll see if I can throw something together to give you an idea | 14:08.02 |
| deekej | chrisl_laptop: I don't want you to burn your time on this :) (and I haven't had some nice challenge lately, it's good to excercise my brain a little ;)) | 14:08.43 |
| | anyway, I'm checking this option for now: http://gallium.inria.fr/blog/portable-conditionals-in-makefiles/ | 14:08.52 |
| | it does not use conditionals at all, but might be usable for this | 14:09.15 |
| | (by conditionals I meant the regular if statements) | 14:09.43 |
| chrisl_laptop | I think that's a similar trick to what we do elsewhere. I would prefer this done directly by susbsitution from configure. | 14:10.19 |
| deekej | okay, I can check that :) | 14:11.05 |
| chrisl_laptop | deekej: the top Makefile is the one that people sometimes have to look at and tweak, so I'd rather keep the nastier tricks out of it. | 14:14.29 |
| deekej | chrisl_laptop: okay, I'll try to come up with something that meets this criteria :) | 14:20.15 |
| | chrisl_laptop: I think I have it. I'm testing it now. | 15:02.33 |
| | seems like it is working as intended | 15:10.42 |
| | I will polish the patch and post it to you | 15:10.54 |
| terryd | my problem is pretty simple, I'm using ghostscript with scribus and unable to import an eps. what i see is 'error: undefined in DELAYBIND' | 23:57.21 |
| | Forward 1 day (to 2017/11/30)>>> | |
Log of #ghostscript at irc.freenode.net.