>>>>> "Angus" == Angus Leeming <[EMAIL PROTECTED]> writes:
Angus> However, I realize that mine is not the only opinion and that
Angus> the 1.3.x tree is Jean-Marc's dominion. So, Jean-Marc, should I
Angus> leave this functionality as-is, or may I abort() if the
Angus> user_lyxdir cannot be created?
I think indeed that aborting is the right thing to do.
Angus> 2. You'll see that gettext is now initialised with the location
Angus> of the localedir in lyx_main.C, the value of this path having
Angus> been established in package.C. This gettext initialisation
Angus> overrides an earlier initialisation in main.C. Is this earlier
Angus> initialisation now redundant, or should I leave the code as-is?
It was an effort to get localization up and running as soon as
possible. For example, to localize the output of -help.
Angus> The final merger will get rid of GetEnvPath also. It's
Angus> currently used in main.C: string lyx_localedir =
Angus> GetEnvPath("LYX_LOCALEDIR"); os_os2.C: string sh =
Angus> OnlyFilename(GetEnvPath("EMXSHELL")); sh =
Angus> OnlyFilename(GetEnvPath("COMSPEC"));
Angus> Clearly, how I get rid of it (completely, or replace it with
Angus> vector<string> getEnvPath()) depends on the answer to question
Angus> (2) above.
LYX_LOCALEDIR is a single directory anyway, so you can use a plain
GetEnv+os::internal_path.
Angus> Finally, whilst I'm asking for opinions, I'd like to propose
Angus> changing the names of these directories:
Since the command line options are -userdir and -sysdir, I'd stick to
something which is not too different.
Angus> It seems to me that "lyx::package().system_support_dir()" is
Angus> rather more elegant than "lyx::package().system_lyxdir()". If
Angus> no one objects, I'll just do it.
I do not really care, actually.
JMarc
