Re: t0021-conversion: flaky filter process test
- From
Johannes Schindelin <johannes.schindelin@gmx.de>
- Date
- Jan 15, 2019, 13:44 UTC
- Message-ID
- <nycvar.QRO.7.76.6.1901151442420.41@tvgsbejvaqbjf.bet>
- In-Reply-To
- <20190111140408.GC840@szeder.dev>
Hi,
On Fri, 11 Jan 2019, SZEDER Gábor wrote:
Show 11 quoted lines
> [...] > > Now, in our tests commands are executed in rapid succession, so it's > highly probable that those files are racily clean, and, consequently, > 'git checkout' has to look at the contents of the files, meaning that > it has to run the clean filter process, and then all is well: that's > why the test succeeds most of the time. Occasionally, however, the > commands are executed slow enough or just at the right moment that the > files are really clean, and 'git checkout' won't bother starting the > clean filter process, which then won't write a 'debug.log' file, > resulting in the above failure.
We had a similar problem in some of the VFS4Git tests, and solved it by using `test-tool chmtime -60 .git/index`. Maybe the same strategy works here?
Ciao, Dscho