On 23 May 2017 at 16:59, Bruce Dubbs <[email protected]> wrote: > Richard Melville wrote: > >> On 22 May 2017 at 19:56, Bruce Dubbs <[email protected] >> <mailto:[email protected]>> wrote: >> >> Richard Melville wrote: >> >> On 22 May 2017 at 18:18, Bruce Dubbs <[email protected] >> <mailto:[email protected]> >> <mailto:[email protected] <mailto:[email protected]>>> >> wrote: >> >> Richard Melville wrote: >> >> >> >> On 21 May 2017 at 16:22, Bruce Dubbs >> <[email protected] <mailto:[email protected]> >> <mailto:[email protected] <mailto: >> [email protected]>> >> <mailto:[email protected] >> <mailto:[email protected]> <mailto:[email protected] >> <mailto:[email protected]>>>> wrote: >> >> Richard Melville wrote: >> >> A minor error but it made my script fail:- >> >> *Installed Directory:* /usr/share/icons/hicolor >> >> This package does not install that directory, >> but rather >> installs >> the file >> "index.theme" in that directory. >> >> >> Checking my log: >> >> BLFS Start INSTALL >> make[1]: Entering directory >> '/tmp/hicolor/hicolor-icon-theme-0.15' >> make[1]: Nothing to be done for 'install-exec-am'. >> /bin/mkdir -p '/usr/share/icons/hicolor' >> /usr/bin/install -c -m 644 index.theme >> '/usr/share/icons/hicolor' >> >> So you are right that does install "index.theme", >> but it does >> also >> create the /usr/share/icons/hicolor directory if it >> does not >> already >> exist. >> >> >> Thanks for your reply Bruce. I think it would be a help >> (it certainly >> would have been for me) if the essence of your reply is >> added as a >> note to >> the hicolor-icon-theme page in the book. As an adjunct >> to this, >> and from >> my discussion with Ken, it's clear that not all package >> build activity >> (creating/writing to directories) is recorded in the >> book, but >> maybe that >> would make the book too dense. I'd be interested in your >> view. >> >> >> I do not understand what your problem is. hicolor-icon-theme >> DOES >> create the directory /usr/share/icons/hicolor as demonstrated >> in the >> mkdir command in the log. >> >> >> The key to the issue is as you expressed it in your last post on >> the subject:- >> >> "...but it does also create the /usr/share/icons/hicolor directory >> *if it >> does not already >> exist.*" My emphasis. >> >> In other words, if /usr/share/icons/hicolor does *already* exist >> then >> hicolor-icon-theme does *not* create it, it merely adds >> index.theme to >> it. On my build, amongst many other packages, the order of build >> was >> emacs-25.1, djvulibre-3.5.27, hicolor-icon-theme-0.15. My build >> logs >> clearly show that emacs created and wrote to >> /usr/share/icons/hicolor, >> then djvulibre created further files in that directory, and >> finally, >> hicolor-icon-theme added index.theme to the already existing >> directory tree. >> >> It's not a build problem as such, merely misleading information >> on the >> hicolor-icon-theme page in the BLFS book. It's misleading because >> it is >> stated clearly that hicolor-icon-theme creates >> /usr/share/icons/hicolor >> when on my build, and probably many others, it didn't. Maybe my >> scripts >> are arcane, but they always look for an installed >> directory/file/header to >> see whether that package has already been installed or not. In >> this case >> the script saw that /usr/share/icons/hicolor already existed and >> assumed >> that hicolor-icon-theme had already been installed when, in >> reality, it >> hadn't. >> >> All I'm suggesting is an addition to the page stating exactly >> what you >> originally confirmed: that hicolor-icon-theme creates >> /usr/share/icons/hicolor ONLY IF IT DOES NOT ALREADY EXIST. >> >> >> I'm not going to open that bucket of worms. Almost every packages >> does something like: >> >> mkdir -p /usr/bin >> >> >> Clearly, I wouldn't use /usr/bin as an identifier for a particular >> package, but rather the unique binary file installed under it. >> >> >> The effort to document such minutia does not justify the trivial >> benefit. >> >> >> The point is that I've built literally hundreds of packages using my >> method and the only one to fail, due to other packages trying to create >> the same directory tree, has been hicolor-icon-theme. That's not to say >> that there aren't others, but I haven't come across them yet. It's true >> that, on my build, both emacs and djvulibre tried to create the same >> directory tree but they both have unique bins/libs that can be used as >> identifiers. All I'm asking is that on the hicolor-icon-theme page we add >> "index.theme" to the directory tree that it creates, thus making it >> unique. >> > > index.theme is not a directory. Since you are the only one with this > problem, why don't you just adjust your script to handle an exception?
I realise that index.theme is not a directory, and, of course, I have adjusted my script already. I just felt that the book shouldn't be stating something that's factually incorrect. To labour the point: the book states that /usr/share/icons/hicolor is created by hicolor-icon-theme when on my build (and I'm sure on many other people's builds) it wasn't because one of any number of other packages has already created it. Anyway, it's clear that this is going nowhere so let's just leave it there. Richard
-- http://lists.linuxfromscratch.org/listinfo/blfs-support FAQ: http://www.linuxfromscratch.org/blfs/faq.html Unsubscribe: See the above information page
