Hi, While optimizing our scanmonitord setup for speed I realized an issue with scanimage in batch mode.
We perform many postprocessing steps on scanned files with daemons called from scanmonitord (this version is not in CVS yet because it's our development version). They maintain different queues so shutting down the server can be done anytime without loosing anything. Another advantage is that things can run asynchronously and in parallel, which gives better overall performance when scanning large amounts of pages with multiple scanners and performing CPU intensive operations like OCR on them. But, if I load 50 pages into the ADF, scanmonitord will run scanimage in batch mode and documents are moved to the first queue directory after the whole job went through. That's why I decided to change the way scanimage writes it's output files in batch mode so that when the file appears with it's name, it has been fully written and can be postprocessed immediately. Attached patch is against current CVS. It creates a temp file using (unsecure) mktemp(3) and writes to it, then closes the file by temporarily redirecting stdout to /dev/null and finally renames it to the final name. Would be nice if that one could go into CVS. Regards, Simon -------------- next part -------------- A non-text attachment was scrubbed... Name: sane-backends-scanimage-batch-atomic-output-files.patch Type: text/x-patch Size: 2101 bytes Desc: not available Url : http://lists.alioth.debian.org/pipermail/sane-devel/attachments/20090114/77c31834/attachment.bin
