| Commit message (Collapse) | Author | Age | Files | Lines |
|
|
|
|
|
|
|
|
|
|
|
| |
* the info on a font now contains the checksum (might be needed for
font caching)
* very small optimizations in force mode
* database version bumping
* a new purge option (not taken into consideration yet)
* new mechanism to remove old entries in the database without
rebuilding everything, activated all the time but (for performance
reasons) will be activated only with the purge options in the very near
future.
|
|
|
|
|
| |
It is actually worse than that, I only test against trunk, once
texlive2010 binaries are frozen we will test against it.
|
|
|
|
|
| |
We also don't need to call that code from outside anyway (there is no
point).
|
| |
|
| |
|
|
|
|
| |
Also stronger checks on status
|
|
|
|
|
|
|
| |
- the status is now externalized in another file name (in the same
directory), so that it's not loaded when just asking for a font
- the status is now filename->last-modification-timestamp so that we
don't have to checksum the files (much faster)
|
| |
|
| |
|
| |
|
| |
|
| |
|
| |
|
|
|
|
|
|
| |
Rewrite the database script to be a more or less a luaotfload module (so
it can be invoked, say, by fontspec) and actual file writing to the
update script. Not finished yet.
|
| |
|
|
|
|
|
| |
This way when the file is moved around, its filename in the database
will be updated, as suggested by Elie.
|
|
|
|
|
| |
So that the old file content is not erased, right now if the script
ended unexpectedly, we are left with an empty file.
|
| |
|
|
|
|
|
|
| |
Store checksums of already loaded fonts in the database to avoids
reloading them, this isn't much faster but it avoids reloading the fonts
and the huge memory consumption. Is there a faster way to hash a file?
|
| |
|
| |
|
| |
|
|
|
|
| |
as it will change soon...
|
| |
|
| |
|
|
|
|
| |
+ Ooops, I missed one commit...
|
| |
|
|
|
|
| |
The script is no longer to be run standalone.
|
| |
|
|
|
|
|
|
|
|
|
| |
Now the options are handled the standard getopt_long way, with a hack to
support multiple -v's, and --verbose now takes a required level
argument.
Also, we now use require() which already have built-in error checking
(if it doesn't work on some system, then this is a bug to report).
|
| |
|
|
|
|
| |
I think it's quite OK as it is now... all comments are welcome!
|
|
|
|
|
|
|
|
|
| |
Now we scan the fonts only if we can write in the final file. There is
currently (if I understand correctly) no clean way to check if a folder
is writable or not in Lua: lfs.attributes will give you the string
returned by ls -l, but there is no simple way to know if we can actually
write, as we don't know if we are the owner, or if we have access to the
group...
|
|
|
|
| |
Useful for the --sys option
|
|
|
|
|
|
|
| |
If the texmf fonts are in fc-cache (which is the case in Debian's
TeX Live 2009, there were two entries in the database, one with
the filename and the other with the complete path, they were also
scanned twice.
|
| |
|
|
|
|
| |
Now encoding related code gone.
|
|
|
|
|
| |
Now many of our fixes are incorporated upstream or fixed in a different
way.
|
| |
|
| |
|
| |
|
|
|
|
| |
Make sure |kind| is defined before calling |string.lower()| on it.
|
| |
|
|
|
|
|
| |
To minimize differences between original ConTeXt files, this module
overrides functions defined in |luat-dum| module.
|
|
|
|
| |
+ Remove some, now unneeded, code.
|
| |
|
| |
|
| |
|
|
|
|
|
|
| |
Due to a bug in current luatex, some fonts would lack certain required
entries in names table, now we'll just ignore such fonts. The real bug
is fixed in luatex trunk anyway.
|
| |
|