[gdal-dev] GDAL 2.3.3 rc2 available

[Marco Atzeri]

Am 15.12.2018 um 09:44 schrieb Even Rouault:
> Marco,
> 
> A few remarks:
> 
> 1) Please build with ./configure --with-proj[=/path/to/proj/prefix]
> so that classic linking is done, and GDAL doesn't have to guess the exact name
> of the shared object. The dlopen() loading of it is deprecated and will be
> completely removed in GDAL 2.5.0

Hi Even,

gdal package on cygwin is built with
--with-proj=/usr/lib --with-python --with-perl --without-jasper
and there is no problem in the build

checking for proj_create in -lproj... yes
checking proj.h usability... yes
checking proj.h presence... yes
checking for proj.h... yes
configure: proj >= 5 detected

> 2) the absence of PROJ during the tests is known to break many tests

Not the case here. The Gdal shared lib is properly linked

$ cygcheck /usr/bin/cyggdal-20.dll | grep proj
     D:\cygwin64T\bin\cygproj-13.dll

Only that test fails.

> 3) Some of the messages you get are expected. Some tests emit warnings /
> errors on purposs and miss silencing them (or sometimes it is not possible to
> silence). With GDAL 2.4 and the refactoring of the autotest suite with pytest,
> such verbosity is eliminated

Noted.

> 4) cygwin is not really "supported", that is we don't have any continuous
> integration for it, and anything not tested doesn't work (that's my mantra).

The definition of "supported" has different mileage for every project;
seldom upstream can test more than few platforms and almost never cygwin
is one. 99% of the time if a project works on vanilla Unix, it works
also for cygwin.

I am the cygwin package maintainer for proj, geos and gdal.
Proj and geos have zero failures, only on gdal I have not
yet spent enough time to clean the autotest failures as it
requires a separate step for running the test suite, that is
always a time consuming effort.

> 5) For the ogr_index errors you get, looking a bit at the case I see code like
> the following in ogr/ogrsf_frmts/mitab/mitab_miffile.cpp
> 
> #ifndef _WIN32
>      /*-----------------------------------------------------------------
>       * On Unix, make sure extension uses the right cases
>       * We do it even for write access because if a file with the same
>       * extension already exists we want to overwrite it.
>       *----------------------------------------------------------------*/
>      TABAdjustFilenameExtension(pszTmpFname);
> #endif
> 
> I suspect that the _WIN32 macro is defined for Cygwin, so the case adjustement
> is not done, whereas it should since Cygwin is probably case sensitive.
> 
> I guess the ogr_mitab_6 crash you get is related since the ogr_index involves
> mapinfo indexes.

I will look on this

> 
> Thanks for your testing,
> 
> Even
> 

Regards
Marco


---
Diese E-Mail wurde von Avast Antivirus-Software auf Viren geprüft.
https://www.avast.com/antivirus