diff options
Diffstat (limited to 'app-arch/par2cmdline/files/par2cmdline-0.8.1-big-endian.patch')
| -rw-r--r-- | app-arch/par2cmdline/files/par2cmdline-0.8.1-big-endian.patch | 70 |
1 files changed, 70 insertions, 0 deletions
diff --git a/app-arch/par2cmdline/files/par2cmdline-0.8.1-big-endian.patch b/app-arch/par2cmdline/files/par2cmdline-0.8.1-big-endian.patch new file mode 100644 index 000000000000..e45f8f3235d1 --- /dev/null +++ b/app-arch/par2cmdline/files/par2cmdline-0.8.1-big-endian.patch @@ -0,0 +1,70 @@ +# backport upstream pr #144, fixes test failure on big endian archs; +# see https://github.com/Parchive/par2cmdline/issues/143 +From 4f3576a314d7169912842ec9dc1e595e61e52653 Mon Sep 17 00:00:00 2001 +From: Michael Nahas <mike@nahas.com> +Date: Tue, 11 Feb 2020 22:42:08 -0600 +Subject: [PATCH] Fix for Github issue #143. Test did not account for + endianness correctly. + +--- + src/reedsolomon_test.cpp | 30 +++++++++++++++++++++++++++--- + 1 file changed, 27 insertions(+), 3 deletions(-) + +diff --git a/src/reedsolomon_test.cpp b/src/reedsolomon_test.cpp +index 1285b3c..c8d26e9 100644 +--- a/src/reedsolomon_test.cpp ++++ b/src/reedsolomon_test.cpp +@@ -433,6 +433,23 @@ int test3() { + + + // Check that the correct constants are being used for Par2 ++ ++//The test pretends there are 10 input blocks ("NUM_IN") and 1 ++//recovery block ("NUM_REC"), each 1024 bytes long ("BUF_SIZE"). These ++//are all stored in data[11][BUF_SIZE], with the input blocks ++//occupying data[0] through data[9] and the recovery block in ++//data[10]. ++ ++//The test zeroes out the input blocks and then writes a 1 into the ++//first location of the first input block, and into the second ++//location of the second input block, etc. It then generates the ++//recovery block using many calls to ReedSolomon. When that happens, ++//those 1s are multiplied by the coefficients for each input block. So ++//the first location of recovery block holds the coefficient for the ++//first input block, the second location has the coefficient for the ++//second input block, etc. Those values are checked against the ++//expected values passed to the function. ++ + template<typename gtype, typename utype> + int test4(int NUM_IN, int *expected_bases) { + //const int NUM_IN = 10; +@@ -448,8 +465,9 @@ int test4(int NUM_IN, int *expected_bases) { + for (int k = 0; k < BUF_SIZE; k++) { + data[i][k] = (u8)0; + } +- // EXCEPT put a 1 in a different place for each file +- ((gtype *)(&(data[i][0])))[i] = (utype) 1; ++ // EXCEPT write a (little endian) 1 in a different place for each file ++ // In the i-th file, it is written into the i-th location ++ data[i][sizeof(utype)*i] = (u8) 1; + } + // zero recovery + for (int j = 0; j < NUM_REC; j++) { +@@ -488,7 +506,13 @@ int test4(int NUM_IN, int *expected_bases) { + // The recovery file has exponent 1 and should + // contain each base to the power 1. + for (int i = 0; i < NUM_IN; i++) { +- int base = (utype) ((gtype *) &(data[NUM_IN+0][0]))[i]; ++ // read little-endian value ++ utype v = 0; ++ for (int byte_index = 0; byte_index < sizeof(utype); byte_index++) { ++ u8 byte = data[NUM_IN+0][sizeof(utype)*i + byte_index]; ++ v |= (((utype)byte) << (byte_index*8)); ++ } ++ int base = v; + if (base != expected_bases[i]) { + cerr << "base at location " << i << " did not match expected." << endl; + cerr << " base = " << base << endl; +-- +2.17.1 + |