On Fri, Jan 14, 2011 at 4:32 PM, Reinhold Kainhofer <reinhold at kainhofer.com> wrote: > Am Freitag, 7. Januar 2011, um 02:34:20 schrieb m. allan noah: >> I took a quick look at the code, and only see a few minor issues at >> first glance. >> >> 1. MC_AutoDetectionTimeout is not configurable at runtime >> 2. the code only calls sanei_usb_init during sane_init. This prevents >> long-running programs like button press daemons from finding >> hotplugged scanners. This was discussed recently on the list. >> 3. Some of your Option Groups could get their names from saneopts.h >> (SANE_NAME_GEOMETRY, TITLE, DESC, etc) >> >> Other than those things, it looks fine to me. > > Okay, here are updated patches (amended, so the whole backend is in one > patch): > http://www.fam.tuwien.ac.at/~reinhold/sane/magicolor_backend_patches/ > > -) Patches 0001 and 0002 are the USB patches you already reviewed (with > copyright notices added as you wanted), so they are ready to be applied. > > -) Patch 0003 fixes a compiler warning that macros like htobe16 (byteorder.h) > etc. are already defined. In that case, don't overwrite the existing > definitions. (Chris Bagwell okay'ed it) > > -) Patch 0004 is the actual backend > > -) Patch 0005 is an alternative to running autoreconf (No manual changes done > by me!). This patch should probably not be applied to git master, but someone > (Chris Bagwell already volunteered) should run autoreconf with the recommended > auto(conf|make|...) versions for sane installed. > > > > Changes to patch 0003 are in detail (compared to the last version you looked > at): > -) All three timeouts (SNMP detection, scan data read, other scanner > responses) are configurable in the magicolor.conf file > -) Cleaned up the options, reused pre-defined values as much as possible > -) Call sanei_usb_init also in get_devices > -) get_devices does not erase the list of scanners any more, but keeps still- > existing scanners untouched, just removes removed and adds new scanners. >
Thank you very much for making those updates. I'll commit the code this weekend, and take a swing at all the autofoo. I'll ask Chris for help if required. > There is, however, one issue that needs to be discussed in principle: > How shall network auto-detection work in a sane way, once several backends > support auto-detecting network-connected scanners? AFAICS, all backends are > loaded sequentially, and the network auto-detection of each backend needs to > wait for a certain time before timing out (even if one response was received, > you have to wait the whole timeout, as there might be another device on the > network). > So, if 10 backends support auto-detection, and each uses a 2 second timeout, > sane would have a startup time of >20 seconds... Clearly, that's not > desirable. > > On Windows, that's not an issue, as the user only installs drivers for the > scanners he really owns. However, sane provides all backends by default and > runs each backend's initialization (get_devices), so here it is a problem. Yes- the SUSE way is to comment out backends in dll.conf. Unfortunately, we dont have a generic way to do that. You should register for an alioth account, and request to join the project, and I will ad you. allan -- "The truth is an offense, but not a sin"
