| // Copyright 2016 The Bazel Authors. All rights reserved. | |
| // | |
| // Licensed under the Apache License, Version 2.0 (the "License"); | |
| // you may not use this file except in compliance with the License. | |
| // You may obtain a copy of the License at | |
| // | |
| // http://www.apache.org/licenses/LICENSE-2.0 | |
| // | |
| // Unless required by applicable law or agreed to in writing, software | |
| // distributed under the License is distributed on an "AS IS" BASIS, | |
| // WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. | |
| // See the License for the specific language governing permissions and | |
| // limitations under the License. | |
| #include "src/tools/singlejar/combiners.h" | |
| #include <cctype> | |
| #include "src/tools/singlejar/diag.h" | |
| Combiner::~Combiner() {} | |
| Concatenator::~Concatenator() {} | |
| bool Concatenator::Merge(const CDH *cdh, const LH *lh) { | |
| if (insert_newlines_ && buffer_.get() && buffer_->data_size() && | |
| '\n' != buffer_->last_byte()) { | |
| Append("\n", 1); | |
| } | |
| CreateBuffer(); | |
| if (Z_NO_COMPRESSION == lh->compression_method()) { | |
| buffer_->ReadEntryContents(lh); | |
| } else if (Z_DEFLATED == lh->compression_method()) { | |
| if (!inflater_) { | |
| inflater_.reset(new Inflater()); | |
| } | |
| buffer_->DecompressEntryContents(cdh, lh, inflater_.get()); | |
| } else { | |
| diag_errx(2, "%s is neither stored nor deflated", filename_.c_str()); | |
| } | |
| return true; | |
| } | |
| void *Concatenator::OutputEntry(bool compress) { | |
| if (!buffer_) { | |
| return nullptr; | |
| } | |
| // Allocate a contiguous buffer for the local file header and | |
| // deflated data. We assume that deflate decreases the size, so if | |
| // the deflater reports overflow, we just save original data. | |
| size_t deflated_buffer_size = | |
| sizeof(LH) + filename_.size() + buffer_->data_size(); | |
| // Huge entry (>4GB) needs Zip64 extension field with 64-bit original | |
| // and compressed size values. | |
| uint8_t | |
| zip64_extension_buffer[sizeof(Zip64ExtraField) + 2 * sizeof(uint64_t)]; | |
| bool huge_buffer = ziph::zfield_needs_ext64(buffer_->data_size()); | |
| if (huge_buffer) { | |
| deflated_buffer_size += sizeof(zip64_extension_buffer); | |
| } | |
| LH *lh = reinterpret_cast<LH *>(malloc(deflated_buffer_size)); | |
| if (lh == nullptr) { | |
| return nullptr; | |
| } | |
| lh->signature(); | |
| lh->version(20); | |
| lh->bit_flag(0x0); | |
| lh->last_mod_file_time(1); // 00:00:01 | |
| lh->last_mod_file_date(30 << 9 | 1 << 5 | 1); // 2010-01-01 | |
| lh->crc32(0x12345678); | |
| lh->compressed_file_size32(0); | |
| lh->file_name(filename_.c_str(), filename_.size()); | |
| if (huge_buffer) { | |
| // Add Z64 extension if this is a huge entry. | |
| lh->uncompressed_file_size32(0xFFFFFFFF); | |
| Zip64ExtraField *z64 = | |
| reinterpret_cast<Zip64ExtraField *>(zip64_extension_buffer); | |
| z64->signature(); | |
| z64->payload_size(2 * sizeof(uint64_t)); | |
| z64->attr64(0, buffer_->data_size()); | |
| lh->extra_fields(reinterpret_cast<uint8_t *>(z64), z64->size()); | |
| } else { | |
| lh->uncompressed_file_size32(buffer_->data_size()); | |
| lh->extra_fields(nullptr, 0); | |
| } | |
| uint32_t checksum; | |
| uint64_t compressed_size; | |
| uint16_t method; | |
| if (compress) { | |
| method = buffer_->CompressOut(lh->data(), &checksum, &compressed_size); | |
| } else { | |
| buffer_->CopyOut(lh->data(), &checksum); | |
| method = Z_NO_COMPRESSION; | |
| compressed_size = buffer_->data_size(); | |
| } | |
| lh->crc32(checksum); | |
| lh->compression_method(method); | |
| if (huge_buffer) { | |
| lh->compressed_file_size32(ziph::zfield_needs_ext64(compressed_size) | |
| ? 0xFFFFFFFF | |
| : compressed_size); | |
| // Not sure if this has to be written in the small case, but it shouldn't | |
| // hurt. | |
| const_cast<Zip64ExtraField *>(lh->zip64_extra_field()) | |
| ->attr64(1, compressed_size); | |
| } else { | |
| // If original data is <4GB, the compressed one is, too. | |
| lh->compressed_file_size32(compressed_size); | |
| } | |
| return reinterpret_cast<void *>(lh); | |
| } | |
| NullCombiner::~NullCombiner() {} | |
| bool NullCombiner::Merge(const CDH * /*cdh*/, const LH * /*lh*/) { | |
| return true; | |
| } | |
| void *NullCombiner::OutputEntry(bool /*compress*/) { return nullptr; } | |
| XmlCombiner::~XmlCombiner() {} | |
| bool XmlCombiner::Merge(const CDH *cdh, const LH *lh) { | |
| if (!concatenator_) { | |
| concatenator_.reset(new Concatenator(filename_, false)); | |
| concatenator_->Append(start_tag_); | |
| concatenator_->Append("\n"); | |
| } | |
| // To ensure xml concatentation is idempotent, read in the entry being added | |
| // and remove the start and end tags if they are present. | |
| TransientBytes bytes_; | |
| if (Z_NO_COMPRESSION == lh->compression_method()) { | |
| bytes_.ReadEntryContents(lh); | |
| } else if (Z_DEFLATED == lh->compression_method()) { | |
| if (!inflater_) { | |
| inflater_.reset(new Inflater()); | |
| } | |
| bytes_.DecompressEntryContents(cdh, lh, inflater_.get()); | |
| } else { | |
| diag_errx(2, "%s is neither stored nor deflated", filename_.c_str()); | |
| } | |
| uint32_t checksum; | |
| char *buf = reinterpret_cast<char *>(malloc(bytes_.data_size())); | |
| // TODO(b/37631490): optimize this to avoid copying the bytes twice | |
| bytes_.CopyOut(reinterpret_cast<uint8_t *>(buf), &checksum); | |
| int start_offset = 0; | |
| if (strncmp(buf, start_tag_.c_str(), start_tag_.length()) == 0) { | |
| start_offset = start_tag_.length(); | |
| } | |
| uint64_t end = bytes_.data_size(); | |
| while (end >= end_tag_.length() && std::isspace(buf[end - 1])) end--; | |
| if (strncmp(buf + end - end_tag_.length(), end_tag_.c_str(), | |
| end_tag_.length()) == 0) { | |
| end -= end_tag_.length(); | |
| } else { | |
| // Leave trailing whitespace alone if we didn't find a match. | |
| end = bytes_.data_size(); | |
| } | |
| concatenator_->Append(buf + start_offset, end - start_offset); | |
| free(buf); | |
| return true; | |
| } | |
| void *XmlCombiner::OutputEntry(bool compress) { | |
| if (!concatenator_) { | |
| return nullptr; | |
| } | |
| concatenator_->Append(end_tag_); | |
| concatenator_->Append("\n"); | |
| return concatenator_->OutputEntry(compress); | |
| } | |
| PropertyCombiner::~PropertyCombiner() {} | |
| bool PropertyCombiner::Merge(const CDH * /*cdh*/, const LH * /*lh*/) { | |
| return false; // This should not be called. | |
| } | |
| ManifestCombiner::~ManifestCombiner() {} | |
| static const char *MULTI_RELEASE = "Multi-Release: true"; | |
| static const size_t MULTI_RELEASE_LENGTH = strlen(MULTI_RELEASE); | |
| static const char *MULTI_RELEASE_PREFIX = "Multi-Release: "; | |
| static const size_t MULTI_RELEASE_PREFIX_LENGTH = strlen(MULTI_RELEASE_PREFIX); | |
| void ManifestCombiner::AppendLine(const std::string &line) { | |
| if (line.find(MULTI_RELEASE_PREFIX, 0, MULTI_RELEASE_PREFIX_LENGTH) != | |
| std::string::npos) { | |
| if (line.find("true", MULTI_RELEASE_PREFIX_LENGTH) != std::string::npos) { | |
| multi_release_ = true; | |
| } else if (line.find("false", MULTI_RELEASE_PREFIX_LENGTH) != | |
| std::string::npos) { | |
| multi_release_ = false; | |
| } | |
| return; | |
| } | |
| concatenator_->Append(line); | |
| if (line[line.size() - 1] != '\n') { | |
| concatenator_->Append("\r\n"); | |
| } | |
| } | |
| bool ManifestCombiner::Merge(const CDH *cdh, const LH *lh) { | |
| TransientBytes bytes_; | |
| if (Z_NO_COMPRESSION == lh->compression_method()) { | |
| bytes_.ReadEntryContents(lh); | |
| } else if (Z_DEFLATED == lh->compression_method()) { | |
| if (!inflater_) { | |
| inflater_.reset(new Inflater()); | |
| } | |
| bytes_.DecompressEntryContents(cdh, lh, inflater_.get()); | |
| } else { | |
| diag_errx(2, "%s is neither stored nor deflated", filename().c_str()); | |
| } | |
| uint32_t checksum; | |
| char *buf = reinterpret_cast<char *>(malloc(bytes_.data_size())); | |
| bytes_.CopyOut(reinterpret_cast<uint8_t *>(buf), &checksum); | |
| const char *line_start = buf; | |
| const char *data_end = buf + bytes_.data_size(); | |
| while (line_start < data_end && line_start[0] != '\r' && | |
| line_start[0] != '\n') { | |
| const char *line_end = strchr(line_start, '\n'); | |
| // Go past return char to point to next line, or to end of data buffer | |
| line_end = line_end != nullptr ? line_end + 1 : data_end; | |
| if (strncmp(line_start, MULTI_RELEASE, MULTI_RELEASE_LENGTH) == 0) { | |
| // The output jar is a MR-JAR if any of the inputs enable the feature. | |
| // Do not check for `Multi-Release: false` here, since inputs shouldn't | |
| // be allowed to override the configuration for other jars on the | |
| // classpath. | |
| multi_release_ = true; | |
| } | |
| line_start = line_end; | |
| } | |
| free(buf); | |
| return true; | |
| } | |
| void *ManifestCombiner::OutputEntry(bool compress) { | |
| if (multi_release_) { | |
| concatenator_->Append(MULTI_RELEASE); | |
| } | |
| concatenator_->Append("\r\n"); | |
| return concatenator_->OutputEntry(compress); | |
| } |