Date | Commit message (Collapse) |
|
Oops, this completely broke things :x
|
|
I prefer GPLv3 nowadays anyways, but continue allowing GPLv2 because
that's what we started with.
|
|
We favor O_NOATIME in a variety of places, use it when we can,
but fall back gracefully when we cannot.
|
|
open() may fail with EPERM on files we do not own.
|
|
HTML sucks, but this was done when I hated HTML less...
|
|
* pcu-fsync: add support for syncfs() on Linux (via -f flag)
|
|
pandoc 1.8 no longer has this, and we don't have HTML
in our documentation anyways.
|
|
Sometimes it is useful to just commit a single filesystem.
This transparently falls back to using sync(2) if syncfs(2)
is not available.
|
|
bogomips.org went on a URL diet last year and lost
the extra bytes.
|
|
|
|
This displays the percentage of the file mmap'ed.
|
|
mmap() may return NULL on success in certain situations not
applicable to us. The proper failure value is MAP_FAILED.
|
|
*sigh*
|
|
Avoid confusing people with a 'sync' command since this is about
fsync-ing individual files.
|
|
This can be useful for calling fsync(2)/fdatasync(2) on
individual files (and/or their containing directories). This
can be used when a system-wide sync(2) is too expensive on
a busy system.
|
|
I doubt we'll ever see sizeof() returning a value larger than
UINT_MAX...
|
|
Due to the bad cast to unsigned long in the parameter before it,
printf was mishandling the bit output as the last parameter.
|
|
|
|
off_t can either be 32-bits or 64-bits wide on popular
architectures in 2009, so we need to pick between strtol and
strtoll. This allows us to (hopefully) deal with files larger
than 2G on 32-bit machines
|
|
For posix_fadvise and mmap, link to opengroup.org since that is
the official specification and not tied to any particular OS.
For mincore, use the kernel.org manpage links since that is
more likely up-to-date than the previous page (and more
canonical of a source).
|
|
|
|
They're already unconditionally defined in compat-util.h,
so avoid cluttering up the build output with them.
|
|
By convention, required arguments are documented by SCREAMING
|
|
|
|
|
|
pandoc can generate stand-alone HTML, an automatic TOC, and
the only Markdown processor I know of capable of generating
manpages.
Markdown is also a more standard syntax than txt2man and
appropriate for generating HTML versions of manpages.
|
|
|
|
* add links to syscall manpages
* include mailing list info
|
|
* Generate homepage using the README
* More git repository information
* Remove references to non-existent pcu(7) manpage
|
|
|
|
|
|
|
|
We don't do things like mlock.
|
|
|
|
|
|
The other 7 bits are undefined at the moment.
|
|
|
|
|
|
Users won't have to mmap entire files if they don't want to
|
|
The page offset is architecture and kernel-dependent, whereas
the byte offset is universally understood.
|
|
Avoid generating unnecessary disk activity.
|
|
I don't plan on making release tarballs "by hand"
any more. cgit/gitweb will just do it for me.
|
|
|
|
|
|
|