Giuseppe Sacco <[EMAIL PROTECTED]> wrote: > Hi Frank, > > Il giorno dom, 15/06/2008 alle 18.25 +0200, Frank Küster ha scritto: >> Frank Küster <[EMAIL PROTECTED]> wrote: > [...] >> > - Whether the list might be extended without a soname change. >> > >> > I fear that currently I cannot rely on the list, so that I would have to >> > check for unsupported names even if the current version of my program >> > can deal with all of them. >> >> But this is still an open point. > > Yes, I think so. Why would you require a soname change when adding page > sizes? Could you please better explain how would you use the library?
We're going to install a hook in /etc/libpaper.d which calls texconfig-sys to set all the TeX system defaults to the libpaper size. >From the Debian point of view, it would be easy to just use paperconf -s to get the sizes, and use them in setting the defaults. However, this has a couple of disadvantages: I might have to round those values (I haven't checked that yet), but what's more important, it would mean to add lots of Debian-specific code where upstream already cares for paper sizes, just not for as many as libpaper provides. Therefore I decided to write a patch to the upstream scripts, namely texconfig, which adds all the sizes we have in Debian, and suggest to them to accept it. After that, all the hook needs to do is query the paper name from libpaper, and call texconfig accordingly - unless I must expect that libpaper gives me a new paper size which I have not yet taken over into texconfig-sys. Since I cannot rely on the list of available papers, I need to add a check whether the return value from paperconf is known to texconfig. Actually I do *not* suggest to bump the soname with every paper added, unless you have planned that anyway. But I'm asking for a clear documentation. Just write somewhere that applications cannot rely on the list in papersize(5), to make it clear. TIA, Frank -- Frank Küster Debian Developer (teTeX/TeXLive) -- To UNSUBSCRIBE, email to [EMAIL PROTECTED] with a subject of "unsubscribe". Trouble? Contact [EMAIL PROTECTED]
