[sword-devel] Xiphos FTBFS with GCC 5.0
karl at kleinpaste.org
Tue Feb 10 17:20:42 MST 2015
On 02/10/2015 01:25 AM, Matěj Cepl wrote:
> this package failed to build because the limit of the
> instantiation depth has been reached.
Considering that I have no environment in which to test this, it's not
very informative. Is there a build log?
Absent a log I can see, I will make an educated guess, that this problem
occurs in modulecache.hh, because clang++ crashes on such a misguided
claim (with absurd/insane values, literally in the tens of thousands).
The depth there is because the cache is instantiated as a layered series
of string- and int-keyed std::maps.
// In precision order.
typedef std::map < int, CacheVerse > CacheChapter;
typedef std::map < int, CacheChapter > CacheBook;
typedef std::map < int, CacheBook > CacheTestament;
typedef std::map < int, CacheTestament > CacheModule;
// Index by module name => Cache for that module.
typedef std::map < const std::string, CacheModule > CacheMap;
// CacheMap contains modules of testaments of books of chapters of
// verses, subscriptable at any stage to get the subordinate content.
However I fail to see how this is a problem, considering that this is in
all honesty a rather bland use of nested maps, and only 5 deep. Also,
the referenced message observes...
> These consist of either package bugs, or GCC bugs.
...and I am inclined to claim that they have a new bug to fix, because
this just shouldn't be a problem, and isn't a problem in gcc4.
Anyhow, I am making suppositions out of air about this; I need a log to
stare at, to see where the problem claimed actually occurred.
Is gcc5 intended to be mandatory in f22?
-------------- next part --------------
An HTML attachment was scrubbed...
More information about the sword-devel