I think you misunderstood. Cygwin is only needed to build the installer, not use/run it.
Workers won't help. It may leave the UI as responsive, but you still have to wait several minutes for the checksum to be completed. MD5 is totally async in the installer today. What isn't is the unzip. It makes the MD5 look like it is stuck, but it isn't. On 6/2/14 12:06 PM, "OmPrakash Muppirala" <bigosma...@gmail.com> wrote: >On Mon, Jun 2, 2014 at 12:00 PM, Alex Harui <aha...@adobe.com> wrote: > >> As you may have seen, I've added a bunch of MD5 checksums to the >>installer >> in hopes of detecting bad downloads and reporting them to the tracker so >> we can see if bad downloads is the root cause of many of the install >> failures that have been reported. >> >> The current built-in MD5Stream, written in AS, is extremely slow for >>large >> downloads like the AIR SDK. On my Mac, the CPU is up around 95% or even >> 100% and still it takes several minutes. I spent the weekend figuring >>out >> how to do the MD5 in CrossBridge. It takes only a few seconds to >>compute >> the checksum. >> >> I think it is worth adding CrossBridge as a build prerequisite to the >> Installer. This will not affect SDK builds. Note that, for Windows >>builds >> of the installer, it will require Cygwin. What do others think? >> >> Thanks, >> -Alex >> >> >-1 for Cygwin requirement on Windows. Cygwin is a pain to setup and >maintain and we definitely dont want end users to jump through this >barrier >to get up and running with Apache Flex. It would defeat the whole point >of >the ease of use of the Installer. > >Couple of options I can think of: >1. Use a NativeProcess like we use for untar on Macs? We can ship the >native process file along with the AIR installer if needed. > >2. Use AS3 Workers to do the MD5 check while still maintaining the UI >responsive. > >3. One thing I always wanted to trouble shoot was - the current MD5 util >we are using is supposed to be async, i.e. we can keep calculating the MD5 >while the file is getting downloaded. For some reason, that does not seem >to be happening. Maybe you can check what is going on there? > >Thanks, >Om
