diff --git a/base/inference.jl b/base/inference.jl
index 432f1ac25ab74..dedbf6f70bd5d 100644
--- a/base/inference.jl
+++ b/base/inference.jl
@@ -267,6 +267,8 @@ const tupleref_tfunc = function (A, t, i)
             end
         elseif i == n && vararg
             return last.parameters[1]
+        elseif i <= 0
+            return None
         else
             return tupleref(t,i)
         end
diff --git a/base/linalg/factorization.jl b/base/linalg/factorization.jl
index 40a48a6fec1b4..a05df7a19509a 100644
--- a/base/linalg/factorization.jl
+++ b/base/linalg/factorization.jl
@@ -10,8 +10,14 @@ macro assertnonsingular(A, info)
    :(($info)==0 ? $A : throw(SingularException($info)))
 end
 
-\(F::Factorization, b::Union(AbstractVector, AbstractMatrix)) = A_ldiv_B!(F, copy(b))
-A_ldiv_B!(C::Factorization, B::StridedVecOrMat) = A_ldiv_B!(C, float(B))
+typealias AbstractVecOrMat Union(AbstractVector, AbstractMatrix)
+
+\(F::Factorization, B::AbstractVecOrMat) = A_ldiv_B!(F, copy(B))
+Ac_ldiv_B(F::Factorization, B::AbstractVecOrMat) = Ac_ldiv_B!(F, copy(B))
+At_ldiv_B(F::Factorization, B::AbstractVecOrMat) = At_ldiv_B!(F, copy(B))
+A_ldiv_B!(F::Factorization, B::StridedVecOrMat) = A_ldiv_B!(F, float(B))
+Ac_ldiv_B!(F::Factorization, B::StridedVecOrMat) = Ac_ldiv_B!(F, float(B))
+At_ldiv_B!(F::Factorization, B::StridedVecOrMat) = At_ldiv_B!(F, float(B))
 
 type Cholesky{T<:BlasFloat} <: Factorization{T}
     UL::Matrix{T}
diff --git a/base/linalg/sparse.jl b/base/linalg/sparse.jl
index 0cc22f9f15769..4946b20e121ca 100644
--- a/base/linalg/sparse.jl
+++ b/base/linalg/sparse.jl
@@ -126,6 +126,62 @@ function *{Tv,Ti}(A::SparseMatrixCSC{Tv,Ti}, B::SparseMatrixCSC{Tv,Ti})
     Ctt = Base.SparseMatrix.transpose!(Ct, SparseMatrixCSC(mA, nB, colptrC, rowvalC, nzvalC))
 end
 
+## solvers
+function A_ldiv_B!(A::SparseMatrixCSC, b::AbstractVector)
+    if istril(A) 
+        if istriu(A) return A_ldiv_B!(Diagonal(A.nzval), b) end
+        return fwdTriSolve!(A, b) 
+    end
+    if istriu(A) return bwdTriSolve!(A, b) end
+    return A_ldiv_B!(lufact(A),b)
+end
+
+function fwdTriSolve!(A::SparseMatrixCSC, b::AbstractVector)
+# forward substitution for CSC matrices
+    n = length(b)
+    ncol = chksquare(A)
+    if n != ncol throw(DimensionMismatch("A is $(ncol)X$(ncol) and b has length $(n)")) end
+   
+    aa = A.nzval
+    ja = A.rowval
+    ia = A.colptr
+   
+    for j = 1:n - 1
+        i1 = ia[j]
+        i2 = ia[j+1]-1
+        b[j] /= aa[i1]
+        bj = b[j]
+        for i = i1+1:i2
+            b[ja[i]] -= bj*aa[i]
+        end
+    end
+    b[end] /= aa[end]
+    return b
+end
+
+function bwdTriSolve!(A::SparseMatrixCSC, b::AbstractVector)
+# backward substitution for CSC matrices
+    n = length(b)
+    ncol = chksquare(A)
+    if n != ncol throw(DimensionMismatch("A is $(ncol)X$(ncol) and b has length $(n)")) end
+
+   aa = A.nzval
+   ja = A.rowval
+   ia = A.colptr
+
+   for j = n:-1:2
+      i1 = ia[j]
+      i2 = ia[j+1]-1
+      b[j] /= aa[i2]
+      bj = b[j]
+      for i = i2-1:-1:i1
+         b[ja[i]] -= bj*aa[i]
+      end
+   end
+   b[1] /= aa[1]
+   return b
+end
+
 ## triu, tril
 
 function triu{Tv,Ti}(S::SparseMatrixCSC{Tv,Ti}, k::Integer)
diff --git a/base/linalg/umfpack.jl b/base/linalg/umfpack.jl
index ae5d26775a798..9f6289a1a43ea 100644
--- a/base/linalg/umfpack.jl
+++ b/base/linalg/umfpack.jl
@@ -8,7 +8,7 @@ export UmfpackLU,
 
 import Base: (\), Ac_ldiv_B, At_ldiv_B, findnz, getindex, nnz, show, size
 
-import ..LinAlg: Factorization, det, lufact, lufact!, solve
+import ..LinAlg: A_ldiv_B!, Ac_ldiv_B!, At_ldiv_B!, Factorization, det, lufact, lufact!, solve
 
 include("umfpack_h.jl")
 type MatrixIllConditionedException <: Exception
@@ -278,42 +278,29 @@ for (sym_r,sym_c,num_r,num_c,sol_r,sol_c,det_r,det_z,lunz,get_num_r,get_num_z,it
 end
 
 ### Solve with Factorization
-
-(\){T<:UMFVTypes}(fact::UmfpackLU{T}, b::Vector{T}) = solve(fact, b)
-(\){Ts<:UMFVTypes,Tb<:Number}(fact::UmfpackLU{Ts}, b::Vector{Tb}) = fact\convert(Vector{Ts},b)
-function (\){Tb<:Complex}(fact::UmfpackLU{Float64}, b::Vector{Tb})
-    r = fact\[convert(Float64,real(be)) for be in b]
-    i = fact\[convert(Float64,imag(be)) for be in b]
-    Complex128[r[k]+im*i[k] for k = 1:length(r)]
-end
-At_ldiv_B{T<:UMFVTypes}(fact::UmfpackLU{T}, b::Vector{T}) = solve(fact, b, UMFPACK_Aat)
-At_ldiv_B{Ts<:UMFVTypes,Tb<:Number}(fact::UmfpackLU{Ts}, b::Vector{Tb}) = fact.'\convert(Vector{Ts},b)
-At_ldiv_B{Tb<:Complex}(fact::UmfpackLU{Float64}, b::Vector{Tb}) = fact.'\b
-Ac_ldiv_B{T<:UMFVTypes}(fact::UmfpackLU{T}, b::Vector{T}) = solve(fact, b, UMFPACK_At)
-Ac_ldiv_B{Ts<:UMFVTypes,Tb<:Number}(fact::UmfpackLU{Ts}, b::Vector{Tb}) = fact'\convert(Vector{Ts},b)
-Ac_ldiv_B{Tb<:Complex}(fact::UmfpackLU{Float64}, b::Vector{Tb}) = fact'\b
-
-### Solve directly with matrix
-
-(\)(S::SparseMatrixCSC, b::Vector) = lufact(S) \ b
-At_ldiv_B{T<:UMFVTypes}(S::SparseMatrixCSC{T}, b::Vector{T}) = solve(lufact(S), b, UMFPACK_Aat)
-At_ldiv_B{Ts<:UMFVTypes,Tb<:Number}(S::SparseMatrixCSC{Ts}, b::Vector{Tb}) = At_ldiv_B(S, convert(Vector{Ts}, b))
-function At_ldiv_B{Tb<:Complex}(S::SparseMatrixCSC{Float64}, b::Vector{Tb})
-    F = lufact(S)
-    r = solve(F, [convert(Float64,real(be)) for be in b], UMFPACK_Aat)
-    i = solve(F, [convert(Float64,imag(be)) for be in b], UMFPACK_Aat)
-    Complex128[r[k]+im*i[k] for k = 1:length(r)]
+A_ldiv_B!{T<:UMFVTypes}(lu::UmfpackLU{T}, b::Vector{T}) = solve(lu, b, UMFPACK_A)
+function A_ldiv_B!{Tlu<:Real,Tb<:Complex}(lu::UmfpackLU{Tlu}, b::Vector{Tb})
+    r = solve(lu, [convert(Tlu,real(be)) for be in b], UMFPACK_A)
+    i = solve(lu, [convert(Tlu,imag(be)) for be in b], UMFPACK_A)
+    Tb[r[k]+im*i[k] for k = 1:length(r)]
 end
-Ac_ldiv_B{T<:UMFVTypes}(S::SparseMatrixCSC{T}, b::Vector{T}) = solve(lufact(S), b, UMFPACK_At)
-Ac_ldiv_B{Ts<:UMFVTypes,Tb<:Number}(S::SparseMatrixCSC{Ts}, b::Vector{Tb}) = Ac_ldiv_B(S, convert(Vector{Ts}, b))
-function Ac_ldiv_B{Tb<:Complex}(S::SparseMatrixCSC{Float64}, b::Vector{Tb})
-    F = lufact(S)
-    r = solve(F, [convert(Float64,real(be)) for be in b], UMFPACK_At)
-    i = solve(F, [convert(Float64,imag(be)) for be in b], UMFPACK_At)
-    Complex128[r[k]+im*i[k] for k = 1:length(r)]
+A_ldiv_B!{Tlu<:UMFVTypes,Tb<:Number}(lu::UmfpackLU{Tlu}, b::Vector{Tb}) = A_ldiv_B!(lu, convert(Vector{Tlu}, b))
+
+Ac_ldiv_B!{T<:UMFVTypes}(lu::UmfpackLU{T}, b::Vector{T}) = solve(lu, b, UMFPACK_At)
+function Ac_ldiv_B!{Tlu<:Real,Tb<:Complex}(lu::UmfpackLU{Tlu}, b::Vector{Tb})
+    r = solve(lu, [convert(Float64,real(be)) for be in b], UMFPACK_At)
+    i = solve(lu, [convert(Float64,imag(be)) for be in b], UMFPACK_At)
+    Tb[r[k]+im*i[k] for k = 1:length(r)]
 end
+Ac_ldiv_B!{Tlu<:UMFVTypes,Tb<:Number}(lu::UmfpackLU{Tlu}, b::Vector{Tb}) = Ac_ldiv_B!(lu, convert(Vector{Tlu}, b))
 
-solve(lu::UmfpackLU, b::Vector) = solve(lu, b, UMFPACK_A)
+At_ldiv_B!{T<:UMFVTypes}(lu::UmfpackLU{T}, b::Vector{T}) = solve(lu, b, UMFPACK_Aat)
+function At_ldiv_B!{Tlu<:Real,Tb<:Complex}(lu::UmfpackLU{Tlu}, b::Vector{Tb})
+    r = solve(lu, [convert(Float64,real(be)) for be in b], UMFPACK_Aat)
+    i = solve(lu, [convert(Float64,imag(be)) for be in b], UMFPACK_Aat)
+    Tb[r[k]+im*i[k] for k = 1:length(r)]
+end
+At_ldiv_B!{Tlu<:UMFVTypes,Tb<:Number}(lu::UmfpackLU{Tlu}, b::Vector{Tb}) = At_ldiv_B!(lu, convert(Vector{Tlu}, b))
 
 function getindex(lu::UmfpackLU, d::Symbol)
     L,U,p,q,Rs = umf_extract(lu)
diff --git a/base/profile.jl b/base/profile.jl
index 5aa4d08c3c48c..31f769deae4f2 100644
--- a/base/profile.jl
+++ b/base/profile.jl
@@ -22,13 +22,13 @@ end
 #### User-level functions
 ####
 function init(n::Integer, delay::Float64)
-    status = ccall(:profile_init, Cint, (Csize_t, Uint64), n, iround(10^9*delay))
+    status = ccall(:jl_profile_init, Cint, (Csize_t, Uint64), n, iround(10^9*delay))
     if status == -1
         error("could not allocate space for ", n, " instruction pointers")
     end
 end
 
-clear() = ccall(:profile_clear_data, Void, ())
+clear() = ccall(:jl_profile_clear_data, Void, ())
 
 function print{T<:Unsigned}(io::IO = STDOUT, data::Vector{T} = fetch(); format = :tree, C = false, combine = true, cols = Base.tty_cols())
     if format == :tree
@@ -75,17 +75,17 @@ isequal(a::LineInfo, b::LineInfo) = a.line == b.line && a.func == b.func && a.fi
 hash(li::LineInfo) = bitmix(hash(li.func), bitmix(hash(li.file), hash(li.line)))
 
 # C wrappers
-start_timer() = ccall(:profile_start_timer, Cint, ())
+start_timer() = ccall(:jl_profile_start_timer, Cint, ())
 
-stop_timer() = ccall(:profile_stop_timer, Void, ())
+stop_timer() = ccall(:jl_profile_stop_timer, Void, ())
 
-is_running() = bool(ccall(:profile_is_running, Cint, ()))
+is_running() = bool(ccall(:jl_profile_is_running, Cint, ()))
 
-get_data_pointer() = convert(Ptr{Uint}, ccall(:profile_get_data, Ptr{Uint8}, ()))
+get_data_pointer() = convert(Ptr{Uint}, ccall(:jl_profile_get_data, Ptr{Uint8}, ()))
 
-len_data() = convert(Int, ccall(:profile_len_data, Csize_t, ()))
+len_data() = convert(Int, ccall(:jl_profile_len_data, Csize_t, ()))
 
-maxlen_data() = convert(Int, ccall(:profile_maxlen_data, Csize_t, ()))
+maxlen_data() = convert(Int, ccall(:jl_profile_maxlen_data, Csize_t, ()))
 
 function lookup(ip::Uint, doCframes::Bool)
     info = ccall(:jl_lookup_code_address, Any, (Ptr{Void}, Bool), ip, doCframes)
diff --git a/src/julia.expmap b/src/julia.expmap
index e1352c991ff17..6337dcebe6f51 100644
--- a/src/julia.expmap
+++ b/src/julia.expmap
@@ -257,14 +257,14 @@
     jl_lookup_code_address;
     jl_read_sonames;
     rec_backtrace;
-    profile_init;
-    profile_get_data;
-    profile_len_data;
-    profile_maxlen_data;
-    profile_clear_data;
-    profile_start_timer;
-    profile_stop_timer;
-    profile_is_running;
+    jl_profile_init;
+    jl_profile_get_data;
+    jl_profile_len_data;
+    jl_profile_maxlen_data;
+    jl_profile_clear_data;
+    jl_profile_start_timer;
+    jl_profile_stop_timer;
+    jl_profile_is_running;
     jl_reshape_array;
     jl_restore_system_image;
     jl_run_event_loop;
diff --git a/src/profile.c b/src/profile.c
index 2fd8c0b8432be..a7e1f1e838c88 100644
--- a/src/profile.c
+++ b/src/profile.c
@@ -57,7 +57,7 @@ static DWORD WINAPI profile_bt( LPVOID lparam )
     hBtThread = 0;
     return 0;
 }
-DLLEXPORT int profile_start_timer(void)
+DLLEXPORT int jl_profile_start_timer(void)
 {
     running = 1;
     if (hBtThread == 0) {
@@ -78,7 +78,7 @@ DLLEXPORT int profile_start_timer(void)
     }
     return (hBtThread != NULL ? 0 : -1);
 }
-DLLEXPORT void profile_stop_timer(void)
+DLLEXPORT void jl_profile_stop_timer(void)
 {
     running = 0;
 }
@@ -229,7 +229,7 @@ void *mach_profile_listener(void *arg)
     }
 }
 
-DLLEXPORT int profile_start_timer(void)
+DLLEXPORT int jl_profile_start_timer(void)
 {
     kern_return_t ret;
     if (!profile_started) {
@@ -268,7 +268,7 @@ DLLEXPORT int profile_start_timer(void)
     return 0;
 }
 
-DLLEXPORT void profile_stop_timer(void)
+DLLEXPORT void jl_profile_stop_timer(void)
 {
     running = 0;
 }
@@ -296,7 +296,7 @@ static void profile_bt(int dummy)
     }
 }
 
-DLLEXPORT int profile_start_timer(void)
+DLLEXPORT int jl_profile_start_timer(void)
 {
     timerprof.it_interval.tv_sec = 0;
     timerprof.it_interval.tv_usec = 0;
@@ -311,7 +311,7 @@ DLLEXPORT int profile_start_timer(void)
     return 0;
 }
 
-DLLEXPORT void profile_stop_timer(void)
+DLLEXPORT void jl_profile_stop_timer(void)
 {
     running = 0;
 }
@@ -343,7 +343,7 @@ static void profile_bt(int signal, siginfo_t *si, void *uc)
     }
 }
 
-DLLEXPORT int profile_start_timer(void)
+DLLEXPORT int jl_profile_start_timer(void)
 {
     struct sigevent sigprof;
     struct sigaction sa;
@@ -376,7 +376,7 @@ DLLEXPORT int profile_start_timer(void)
     return 0;
 }
 
-DLLEXPORT void profile_stop_timer(void)
+DLLEXPORT void jl_profile_stop_timer(void)
 {
     if (running)
         timer_delete(timerprof);
@@ -389,7 +389,7 @@ DLLEXPORT void profile_stop_timer(void)
 ///////////////////////
 // Utility functions //
 ///////////////////////
-DLLEXPORT int profile_init(size_t maxsize, u_int64_t delay_nsec)
+DLLEXPORT int jl_profile_init(size_t maxsize, u_int64_t delay_nsec)
 {
     bt_size_max = maxsize;
     nsecprof = delay_nsec;
@@ -402,27 +402,27 @@ DLLEXPORT int profile_init(size_t maxsize, u_int64_t delay_nsec)
     return 0;
 }
 
-DLLEXPORT u_int8_t* profile_get_data(void)
+DLLEXPORT u_int8_t* jl_profile_get_data(void)
 {
     return (u_int8_t*) bt_data_prof;
 }
 
-DLLEXPORT size_t profile_len_data(void)
+DLLEXPORT size_t jl_profile_len_data(void)
 {
     return bt_size_cur;
 }
 
-DLLEXPORT size_t profile_maxlen_data(void)
+DLLEXPORT size_t jl_profile_maxlen_data(void)
 {
     return bt_size_max;
 }
 
-DLLEXPORT void profile_clear_data(void)
+DLLEXPORT void jl_profile_clear_data(void)
 {
     bt_size_cur = 0;
 }
 
-DLLEXPORT int profile_is_running(void)
+DLLEXPORT int jl_profile_is_running(void)
 {
     return running;
 }
diff --git a/test/core.jl b/test/core.jl
index 9d8664f044790..60987d84a3a22 100644
--- a/test/core.jl
+++ b/test/core.jl
@@ -1165,3 +1165,14 @@ macro myassert4873(ex)
 end
 x4873 = 1
 @myassert4873 (x -> x)(x4873) == 1
+
+# issue from IRC
+function invalid_tupleref()
+    A = (1, "2", 3.0)
+    try
+        return A[0]
+    catch
+        return true
+    end
+end
+@test invalid_tupleref()==true
diff --git a/test/sparse.jl b/test/sparse.jl
index f7344e9cb3bea..73aac32eba551 100644
--- a/test/sparse.jl
+++ b/test/sparse.jl
@@ -75,12 +75,57 @@ for i = 1:5
     @test (maximum(abs(a*b - full(a)*b)) < 100*eps())
     @test (maximum(abs(a\b - full(a)\b)) < 1000*eps())
     @test (maximum(abs(a'\b - full(a')\b)) < 1000*eps())
+    @test (maximum(abs(a.'\b - full(a.')\b)) < 1000*eps())
+    
     a = speye(5) + 0.1*sprandn(5, 5, 0.2) + 0.1*im*sprandn(5, 5, 0.2)
     b = randn(5)
     @test (maximum(abs(a*b - full(a)*b)) < 100*eps())
     @test (maximum(abs(a\b - full(a)\b)) < 1000*eps())
     @test (maximum(abs(a'\b - full(a')\b)) < 1000*eps())
     @test (maximum(abs(a.'\b - full(a.')\b)) < 1000*eps())
+
+    a = speye(5) + tril(0.1*sprandn(5, 5, 0.2))
+    b = randn(5) + im*randn(5)
+    @test (maximum(abs(a*b - full(a)*b)) < 100*eps())
+    @test (maximum(abs(a\b - full(a)\b)) < 1000*eps())
+    @test (maximum(abs(a'\b - full(a')\b)) < 1000*eps())
+    @test (maximum(abs(a.'\b - full(a.')\b)) < 1000*eps())
+
+    a = speye(5) + tril(0.1*sprandn(5, 5, 0.2) + 0.1*im*sprandn(5, 5, 0.2))
+    b = randn(5)
+    @test (maximum(abs(a*b - full(a)*b)) < 100*eps())
+    @test (maximum(abs(a\b - full(a)\b)) < 1000*eps())
+    @test (maximum(abs(a'\b - full(a')\b)) < 1000*eps())
+    @test (maximum(abs(a.'\b - full(a.')\b)) < 1000*eps())
+
+    a = speye(5) + triu(0.1*sprandn(5, 5, 0.2))
+    b = randn(5) + im*randn(5)
+    @test (maximum(abs(a*b - full(a)*b)) < 100*eps())
+    @test (maximum(abs(a\b - full(a)\b)) < 1000*eps())
+    @test (maximum(abs(a'\b - full(a')\b)) < 1000*eps())
+    @test (maximum(abs(a.'\b - full(a.')\b)) < 1000*eps())
+
+    a = speye(5) + triu(0.1*sprandn(5, 5, 0.2) + 0.1*im*sprandn(5, 5, 0.2))
+    b = randn(5)
+    @test (maximum(abs(a*b - full(a)*b)) < 100*eps())
+    @test (maximum(abs(a\b - full(a)\b)) < 1000*eps())
+    @test (maximum(abs(a'\b - full(a')\b)) < 1000*eps())
+    @test (maximum(abs(a.'\b - full(a.')\b)) < 1000*eps())
+
+    a = speye(5) + triu(0.1*sprandn(5, 5, 0.2))
+    b = randn(5) + im*randn(5)
+    @test (maximum(abs(a*b - full(a)*b)) < 100*eps())
+    @test (maximum(abs(a\b - full(a)\b)) < 1000*eps())
+    @test (maximum(abs(a'\b - full(a')\b)) < 1000*eps())
+    @test (maximum(abs(a.'\b - full(a.')\b)) < 1000*eps())
+
+    a = spdiagm(randn(5)) + im*spdiagm(randn(5))
+    b = randn(5)
+    @test (maximum(abs(a*b - full(a)*b)) < 100*eps())
+    @test (maximum(abs(a\b - full(a)\b)) < 1000*eps())
+    @test (maximum(abs(a'\b - full(a')\b)) < 1000*eps())
+    @test (maximum(abs(a.'\b - full(a.')\b)) < 1000*eps())
+
     b = randn(5) + im*randn(5)
     @test (maximum(abs(a*b - full(a)*b)) < 100*eps())
     @test (maximum(abs(a\b - full(a)\b)) < 1000*eps())