Re: [Boost-testing] [graph] Testing modifies sources?

From: Victor A. Wagner Jr. (vawjr_at_[hidden])
Date: 2007-05-27 23:44:55

Rene Rivera wrote:
> Rene Rivera wrote:
>   
>> Any ideas how those end up being changed?
>>     
>
> And here's the reason...
>
> 1. Do a clean checkout, which creates all the files with the current 
> date-time. The lex files now have essentially the same timestamp and CVS 
> sets them to the time when it writes the files.
>   
Do NOT assume that any two will have the same timstamp  Those are NOT 
"touched" to make them the same, it's just the time as which they were 
closed.  If you have large (or many) you will see differences in the 
timestamps.
> 2. Run tests, i.e. bjam, which see the equal
(see above, they're not necessarily equal)
>  timestamps and rebuilds 
> those files because they are not newer than the dependencies.
>
> 3. Do an update, when doing the next test iteration, and one sees the 
> "M" files.
>
> My current solution was to delete the two lex files, and on the next 
> update CVS gives them a timestamp later than the rest of the files. The 
> real solution would be for the version control software to set the 
> timestamps to the one in the repo.
NO...this causes catastrophic problems if you ever have to get an older 
version of something (why have a version control at all if you're never 
going to do this) the source will predate the "object" by quite a bit 
and look exactly up to date.
>  But AFAICT neither CVS, nor SVN, have 
> such an option. Another possible solution is to change bjam to treat 
> equal timestamps as up to date in the dependency analysis.
>
>
>