On Mon, Sep 17, 2018 at 5:36 PM Pavel Stehule <pavel.stehule@gmail.com> wrote: > po 17. 9. 2018 v 2:05 odesílatel Thomas Munro <thomas.munro@enterprisedb.com> napsal: >> On Fri, Aug 10, 2018 at 6:26 AM Andrew Dunstan >> <andrew.dunstan@2ndquadrant.com> wrote: >> > On 01/24/2018 04:30 AM, Pavel Stehule wrote: >> > > >> > > I am sending updated version. >> > > >> > > Very much thanks for very precious review >> > >> > Thomas, >> > >> > I am unable to replicate the Linux failure seen in the cfbot on my >> > Fedora machine. Both when building with libxml2 and without, after >> > applying the latest patch the tests pass without error. Can you please >> > investigate what's going on here? >> >> Well this is strange... I can't reproduce the problem either with or >> without --with-libxml on a Debian box (was trying to get fairly close >> to the OS that Travis runs on). But I see the same failure when I >> apply the patch on my FreeBSD 12 laptop and test without >> --with-libxml. Note that when cfbot runs it, the patch is applied >> with FreeBSD patch, and then it's tested without --with-libxml on >> Ubuntu (Travis's default OS). [Side note: I should change it to build >> --with-libxml, but that's not the point.] So the common factor is a >> different patch implementation. I wonder if a hunk is being >> misinterpreted. > > > This patch is not too large. Please, can me send a related files, I can check it manually.
I confirmed that xml_1.out is different depending on which 'patch' you use. I've attached the output from FreeBSD patch 2.0-12u11 and GNU patch 2.5.8. It's an interesting phenomenon, probably due to having a huge long file with a lot of repeated text and a slightly different algorithms or parameters, but I don't think you need to worry about it for this. Sorry for the distraction.