Subject: Re: [Boost-build] --hash option undocumented?
From: Matthew Chambers (matt.chambers42_at_[hidden])
Date: 2012-10-19 16:12:23
On 10/19/2012 9:45 AM, Steven Watanabe wrote:
> On 10/19/2012 04:58 AM, John Maddock wrote:
>> I've just had a very capricious issue with path names exceeding the
>> system maximum which caused doxygen to produce truncated output (it
>> didn't fail though, it actually looked like an xsltproc failure later).
>> Once I realized what the issue was I knew there was a bjam option to
>> shorten the path names, but I haven't been able to find it in the docs,
>> or in the output of bjam --help. I had to search the mailing to
>> eventually track it down :-(
>> 3) Is there any way that Boost.Build can detect over long paths up front
>> and at least warn you about this issue? It can take an awful long time
>> to track down the cause if you're not expecting it.
> For doxygen, it isn't really possible, because
> Boost.Build doesn't know about all the files
> that doxygen creates.
I posted about this issue earlier, having had the same very capricious issue. So how hard would it
be to check all target paths on Windows for length >= 260 and trigger a fatal error (suggesting
--hash or --abbrev-paths)?
> On Thu, May 19, 2011 at 2:21 AM, Matthew Chambers <matt.chambers42_at_[hidden]
> <mailto:matt.chambers42_at_[hidden]>> wrote:
> On 5/18/2011 11:08 AM, Rene Rivera wrote:
> On 5/18/2011 10:57 AM, Matthew Chambers wrote:
> Having been bitten a few times by this in the middle of a build - at
> which point I had to restart with --hash - it got me thinking that this
> should be tested as much as possible before starting the action phase.
> Just run through all the target output paths and make sure they don't
> exceed MAX_PATH. How hard would this be to implement? It won't catch
> cases inside the actions of custom make targets, but it'd catch almost
> Which MAX_PATH are you referring to? And how, or if, does it correspond to the actual
> allowed max path length of the filesystem used during the build?
> 260 characters for the lose. :(
Boost-Build list run by bdawes at acm.org, david.abrahams at rcn.com, gregod at cs.rpi.edu, cpdaniel at pacbell.net, john at johnmaddock.co.uk