Re: [Help] Re: jellyfish FTBFS on 32bit

[Nilesh Patra <npatra974@gmail.com>]

Hi Andreas,

On Thu, 3 Dec 2020 at 03:04, Andreas Tille <andreas@an3as.eu> wrote:

Control: tags -1 help

Hi,

I guess this bug is relatively easy to fix by some explicit typing - but
I'm lacking the C++ knowledge to find out where.

Attaching a patch for this, it goes past this.

However, the build-time tests fail on 32-bit arches - on exploring the test files, it looks like (with some confidence at least) that its due to large valued numbers in test files, which is (probably) beyond what a 32-bit arch can digest.

Maybe it is sensible to disable tests (both build-time and autopkgtest) for that 32-bit arch, but I leave that up to you to check and decide :-)

Kind regards

Nilesh

--- a/include/jellyfish/rectangular_binary_matrix.hpp
+++ b/include/jellyfish/rectangular_binary_matrix.hpp
@@ -118,7 +118,7 @@
       uint64_t *p = _columns;
       while(*p == 0 && p < _columns + _c)
         ++p;
-      return (p - _columns) == _c;
+      return static_cast<unsigned int>(p - _columns) == _c;
     }
 
     // Randomize the content of the matrix
--- a/include/jellyfish/mer_dna.hpp
+++ b/include/jellyfish/mer_dna.hpp
@@ -693,7 +693,7 @@
 
   char buffer[mer.k() + 1];
   is.read(buffer, mer.k());
-  if(is.gcount() != mer.k())
+  if(static_cast<unsigned int>(is.gcount()) != mer.k())
     goto error;
   buffer[mer.k()] = '\0';
   if(!mer.from_chars(buffer))
--- a/unit_tests/test_mer_dna.cc
+++ b/unit_tests/test_mer_dna.cc
@@ -466,7 +466,7 @@
 
     // Get bits by right-shifting
     typename TypeParam::Type cm(m);
-    for(unsigned int j = 1; j < start; j += 2)
+    for(unsigned long int j = 1; start>=0 && j < static_cast<unsigned int>(start); j += 2)
       cm.shift_right(0); // Shift by 2 bits
     typename TypeParam::Type::base_type y = cm.word(0);
     if(start & 0x1)