Some tests fail with Illegal Instruction caused by an assertion (SIGILL)

[yurivict]

Does this problem persist on the current main?

• I have verified the issue on the current main

Is there an existing issue for this?

• I have searched the existing issues

Current Behavior

96% tests passed, 14 tests failed out of 315

Total Test time (real) =  21.86 sec

The following tests FAILED:
        234 - search/search_collection_test (ILLEGAL)
        236 - search/search_scheme_algorithm_test (ILLEGAL)
        238 - search/search_test (ILLEGAL)
        244 - search/fm_index/fm_index_dna4_test (ILLEGAL)
        245 - search/fm_index/bi_fm_index_dna4_test (ILLEGAL)
        246 - search/fm_index/bi_fm_index_aa27_test (ILLEGAL)
        247 - search/fm_index/bi_fm_index_char_test (ILLEGAL)
        248 - search/fm_index_cursor/fm_index_cursor_test (ILLEGAL)
        249 - search/fm_index_cursor/fm_index_cursor_collection_test (ILLEGAL)
        250 - search/fm_index_cursor/bi_fm_index_cursor_test (ILLEGAL)
        251 - search/fm_index_cursor/bi_fm_index_cursor_collection_test (ILLEGAL)
        252 - search/fm_index_cursor/extend_bitpacked_query_sequence_test (ILLEGAL)
        260 - test/expect_same_type_test (Failed)
        278 - utility/detail/type_name_as_string_test (Failed)

For example .test/search/fm_index_cursor/bi_fm_index_cursor_collection_test crashes here:

230│
231│   // 27.6.2.3.2 Get area:
232│   _LIBCPP_HIDE_FROM_ABI char_type* eback() const { return __binp_; }
233│   _LIBCPP_HIDE_FROM_ABI char_type* gptr() const { return __ninp_; }
234│   _LIBCPP_HIDE_FROM_ABI char_type* egptr() const { return __einp_; }
235│
236│   inline _LIBCPP_HIDE_FROM_ABI_AFTER_V1 void gbump(int __n) { __ninp_ += __n; }
237│
238│   inline _LIBCPP_HIDE_FROM_ABI_AFTER_V1 void setg(char_type* __gbeg, char_type* __gnext, char_type* __gend) {
239├───> _LIBCPP_ASSERT_VALID_INPUT_RANGE(std::__is_valid_range(__gbeg, __gnext), "[gbeg, gnext) must be a valid range");
240│     _LIBCPP_ASSERT_VALID_INPUT_RANGE(std::__is_valid_range(__gbeg, __gend), "[gbeg, gend) must be a valid range");
241│     _LIBCPP_ASSERT_VALID_INPUT_RANGE(std::__is_valid_range(__gnext, __gend), "[gnext, gend) must be a valid range");
242│     __binp_ = __gbeg;
243│     __ninp_ = __gnext;
244│     __einp_ = __gend;
245│   }
246│
247│   // 27.6.2.3.3 Put area:
248│   _LIBCPP_HIDE_FROM_ABI char_type* pbase() const { return __bout_; }
249│   _LIBCPP_HIDE_FROM_ABI char_type* pptr() const { return __nout_; }
250│   _LIBCPP_HIDE_FROM_ABI char_type* epptr() const { return __eout_; }

1747│             pbump64(n);
 1748│             return n;
 1749│         }
 1750│         else
 1751│         {
 1752│             if (epptr() - pbase() == (std::ptrdiff_t)m_ram_file->size() and epptr() == pptr())
 1753│             {
 1754│                 m_ram_file->insert(m_ram_file->end(), s, s + n);
 1755│                 setp(m_ram_file->data(), m_ram_file->data() + m_ram_file->size());
 1756│                 std::ptrdiff_t add = epptr() - pbase();
 1757│                 pbump64(add);
 1758├───────────────> setg(m_ram_file->data(), gptr(), m_ram_file->data() + m_ram_file->size());
 1759│                 return n;
 1760│             }
 1761│             else
 1762│             {
 1763│                 for (std::streamsize i = 0; i < n; ++i)
 1764│                 {
 1765│                     if (traits_type::eq_int_type(sputc(s[i]), traits_type::eof()))
 1766│                     {
 1767│                         return i;
 1768│                     }
 1769│                 }

 3130│ template <typename T>
 3131│ size_t write_member(T const & t, std::ostream & out, seqan3::contrib::sdsl::structure_tree_node * v = nullptr, std::string name = "")
 3132│ {
 3133│     seqan3::contrib::sdsl::structure_tree_node * child = seqan3::contrib::sdsl::structure_tree::add_child(v, name, util::class_name(t));
 3134├───> out.write((char *)&t, sizeof(t));
 3135│     size_t written_bytes = sizeof(t);
 3136│     seqan3::contrib::sdsl::structure_tree::add_size(child, written_bytes);
 3137│     return written_bytes;
 3138│ }
 3139│ template <>
 3140│ inline size_t
 3141│ write_member<std::string>(std::string const & t, std::ostream & out, seqan3::contrib::sdsl::structure_tree_node * v, std::string name)

#0  std::__1::basic_streambuf<char, std::__1::char_traits<char> >::setg[abi:se190107](char*, char*, char*) (this=0x800c0f0a0, __gbeg=0x800c1f210 "\250", __gnext=0x0, __gend=0x800c1f218 "")
    at /usr/include/c++/v1/streambuf:239
#1  0x00000000002d6abd in seqan3::contrib::sdsl::ram_filebuf::xsputn (this=0x800c0f0a0, s=0x7fffffff6820 "\250", n=8) at /usr/ports/biology/seqan3/work/seqan3-3.4.2/include/seqan3/contrib/sdsl-lite.hpp:1758
#2  0x0000000800565796 in std::__1::basic_ostream<char, std::__1::char_traits<char> >::write(char const*, long) () from /lib/libc++.so.1
#3  0x00000000002c6a4d in seqan3::contrib::sdsl::write_member<unsigned long> (t=@0x7fffffff6820: 576460752303423656, out=..., v=0x0, name="")
    at /usr/ports/biology/seqan3/work/seqan3-3.4.2/include/seqan3/contrib/sdsl-lite.hpp:3134
#4  0x00000000002dce48 in seqan3::contrib::sdsl::int_vector<(unsigned char)8>::write_header (size=168, int_width=8 '\b', out=...) at /usr/ports/biology/seqan3/work/seqan3-3.4.2/include/seqan3/contrib/sdsl-lite.hpp:5392
#5  0x00000000002d5bec in seqan3::contrib::sdsl::int_vector<(unsigned char)8>::serialize (this=0x7fffffff6c28, out=..., v=0x0, name="")
    at /usr/ports/biology/seqan3/work/seqan3-3.4.2/include/seqan3/contrib/sdsl-lite.hpp:6260
#6  0x00000000002d4bb7 in seqan3::contrib::sdsl::store_to_file<(unsigned char)8> (v=..., file="@50499_0") at /usr/ports/biology/seqan3/work/seqan3-3.4.2/include/seqan3/contrib/sdsl-lite.hpp:3826
#7  0x00000000002d2477 in seqan3::contrib::sdsl::construct_im<seqan3::contrib::sdsl::csa_wt<seqan3::contrib::sdsl::wt_pc<seqan3::contrib::sdsl::balanced_shape, seqan3::contrib::sdsl::int_vector<(unsigned char)1>, seqan3::con
trib::sdsl::rank_support_v<(unsigned char)1, (unsigned char)1>, seqan3::contrib::sdsl::select_support_scan<(unsigned char)1, (unsigned char)1>, seqan3::contrib::sdsl::select_support_scan<(unsigned char)0, (unsigned char)1>,
seqan3::contrib::sdsl::byte_tree<false> >, 16u, 10000000u, seqan3::contrib::sdsl::sa_order_sa_sampling<(unsigned char)0>, seqan3::contrib::sdsl::isa_sampling<(unsigned char)0>, seqan3::contrib::sdsl::plain_byte_alphabet>, se
qan3::contrib::sdsl::int_vector<(unsigned char)8>&> (idx=..., data=..., num_bytes=0 '\000') at /usr/ports/biology/seqan3/work/seqan3-3.4.2/include/seqan3/contrib/sdsl-lite.hpp:21017
#8  0x00000000002d149a in _ZN6seqan38fm_indexINS_4dna4ELNS_11text_layoutE1ENS_7contrib4sdsl6csa_wtINS4_5wt_pcINS4_14balanced_shapeENS4_10int_vectorILh1EEENS4_14rank_support_vILh1ELh1EEENS4_19select_support_scanILh1ELh1EEENSC
_ILh0ELh1EEENS4_9byte_treeILb0EEEEELj16ELj10000000ENS4_20sa_order_sa_samplingILh0EEENS4_12isa_samplingILh0EEENS4_19plain_byte_alphabetEEEE9constructITkNSt3__16ranges5rangeERNSQ_6vectorINSS_IS1_NSQ_9allocatorIS1_EEEENST_ISV_E
EEEQeqT0_LS2_1EEEvOT_b (this=0x7fffffff7ce8, text=std::vector of length 2 = {...}, reverse=false) at /usr/ports/biology/seqan3/work/seqan3-3.4.2/include/seqan3/search/fm_index/fm_index.hpp:373
#9  0x00000000002d0d65 in _ZN6seqan38fm_indexINS_4dna4ELNS_11text_layoutE1ENS_7contrib4sdsl6csa_wtINS4_5wt_pcINS4_14balanced_shapeENS4_10int_vectorILh1EEENS4_14rank_support_vILh1ELh1EEENS4_19select_support_scanILh1ELh1EEENSC
_ILh0ELh1EEENS4_9byte_treeILb0EEEEELj16ELj10000000ENS4_20sa_order_sa_samplingILh0EEENS4_12isa_samplingILh0EEENS4_19plain_byte_alphabetEEEEC2ITkNSt3__16ranges19bidirectional_rangeERNSQ_6vectorINSS_IS1_NSQ_9allocatorIS1_EEEENS
T_ISV_EEEEEEOT_ (this=0x7fffffff7ce8, text=std::vector of length 2 = {...}) at /usr/ports/biology/seqan3/work/seqan3-3.4.2/include/seqan3/search/fm_index/fm_index.hpp:454
#10 0x00000000002be668 in _ZN6seqan311bi_fm_indexINS_4dna4ELNS_11text_layoutE1ENS_7contrib4sdsl6csa_wtINS4_5wt_pcINS4_14balanced_shapeENS4_10int_vectorILh1EEENS4_14rank_support_vILh1ELh1EEENS4_19select_support_scanILh1ELh1EE
ENSC_ILh0ELh1EEENS4_9byte_treeILb0EEEEELj16ELj10000000ENS4_20sa_order_sa_samplingILh0EEENS4_12isa_samplingILh0EEENS4_19plain_byte_alphabetEEEE9constructITkNSt3__16ranges5rangeERNSQ_6vectorINSS_IS1_NSQ_9allocatorIS1_EEEENST_I
SV_EEEEEEvOT_ (this=0x7fffffffc238, text=std::vector of length 2 = {...}) at /usr/ports/biology/seqan3/work/seqan3-3.4.2/include/seqan3/search/fm_index/bi_fm_index.hpp:120
#11 0x00000000002bdd67 in _ZN6seqan311bi_fm_indexINS_4dna4ELNS_11text_layoutE1ENS_7contrib4sdsl6csa_wtINS4_5wt_pcINS4_14balanced_shapeENS4_10int_vectorILh1EEENS4_14rank_support_vILh1ELh1EEENS4_19select_support_scanILh1ELh1EE

Expected Behavior

n/a

Steps To Reproduce

build + tests

Environment

- Operating system: FreeBSD 15 STABLE
- SeqAn version: 3.4.2
- Compiler: clabg-19

Anything else?

No response

[eseiler]

Confirmed on FreeBSD.
Preliminary patch attached.

What happens is that xsputn (and also overflow) may cause reallocation (via insert and push_back, respectively) of the underlying vector.
In such a case, gptr() is a nullptr (or just stale¹).
FreeBSD validates setg ranges, and [valid_char_ptr, nullptr] is obviously not valid.

¹_{Here, "stale" means that, for [begin,end), end cannot be reached from begin.}

I'll fix it in the SDSL repo and then update our amalgamated header.

seqan3_3415.patch Download

Patch content (click to show)

diff --git a/include/seqan3/contrib/sdsl-lite.hpp b/include/seqan3/contrib/sdsl-lite.hpp
index a82da4ac2..b6c7cf061 100644
--- a/include/seqan3/contrib/sdsl-lite.hpp
+++ b/include/seqan3/contrib/sdsl-lite.hpp
@@ -1600,6 +1600,10 @@ class ram_filebuf : public std::streambuf
 {
 private:
     ram_fs::content_type * m_ram_file = nullptr;
+    std::ptrdiff_t get_pos() const
+    {
+        return (eback() != nullptr && gptr() != nullptr) ? (gptr() - eback()) : 0;
+    }
     void pbump64(std::ptrdiff_t x)
     {
         while (x > std::numeric_limits<int>::max())
@@ -1751,11 +1755,12 @@ public:
         {
             if (epptr() - pbase() == (std::ptrdiff_t)m_ram_file->size() and epptr() == pptr())
             {
+                std::ptrdiff_t const read_pos = get_pos();
                 m_ram_file->insert(m_ram_file->end(), s, s + n);
                 setp(m_ram_file->data(), m_ram_file->data() + m_ram_file->size());
                 std::ptrdiff_t add = epptr() - pbase();
                 pbump64(add);
-                setg(m_ram_file->data(), gptr(), m_ram_file->data() + m_ram_file->size());
+                setg(m_ram_file->data(), m_ram_file->data() + read_pos, m_ram_file->data() + m_ram_file->size());
                 return n;
             }
             else
@@ -1779,11 +1784,12 @@ public:
     {
         if (m_ram_file)
         {
+            std::ptrdiff_t const read_pos = get_pos();
             m_ram_file->push_back(c);
             setp(m_ram_file->data(), m_ram_file->data() + m_ram_file->size());
             std::ptrdiff_t add = epptr() - pbase();
             pbump64(add);
-            setg(m_ram_file->data(), gptr(), m_ram_file->data() + m_ram_file->size());
+            setg(m_ram_file->data(), m_ram_file->data() + read_pos, m_ram_file->data() + m_ram_file->size());
         }
         return traits_type::to_int_type(c);
     }