If there is still an issue about scons startup time (which lead Amanda
to suggest trying to get a scons server / deamon, quoted below) I
think I found some good news:
   scons --interactive
Added by Adam Simpkins about a year ago: http://www.scons.org/CHANGES.txt

>From the docs, I guess --interactive only eliminates parsing of
SConscript files. But the kind of "sconsd" Amanda suggested would
probably keep running "build all" and wait for an event/message (from
your text editors "quick compile" button?) about which source file has
changed. Currently scons probably does md5 (#) checks on hordes of
files for each "build all"?

Maybe md5 is the slow part - has someone tried to disable md5
checksums to see the difference? If filedate is too weak, maybe
(filedate XOR filesize) is good enough to survive without (full) md5
checking?

Or make "sconsd" register with the OS for events about file changes,
and then have some code inside scons (a third alternative to existing
md5 vs filedate) keeps this in memory (maybe using a b-tree, where the
leaves can be specific changed files or a whole folder in case most
files in that folder changed. Windows folder watcher:
http://www.codeproject.com/KB/files/directorychangewatcher.aspx

Searching scons' mail list archives gave very little on optimizing for
speed. Just some talk of optimizing by grouping compiler calls (which
I didn't read through):
http://scons.tigris.org/ds/viewMessage.do?dsForumId=1268&dsMessageId=488275
(So, is scons just fast enough, or that scons users don't do big
projects on slow machines?)


On Nov 14 2008, 8:55 pm, Amanda Walker <[email protected]> wrote:
> I don't know anything about scons internals, but how hard would it be
> to fix that?  Have a "sconsserver" that sat around waiting for
> commands to build (or even watched a set of directories for changes
> and updated its dependency graph on the fly)?
>
> --Amanda
--~--~---------~--~----~------------~-------~--~----~
Chromium Developers mailing list: [email protected] 
View archives, change email options, or unsubscribe: 
    http://groups.google.com/group/chromium-dev
-~----------~----~----~----~------~----~------~--~---

Reply via email to