Hi all, I am writing an implementation of the git HTTP pack protocol in C. It just does a request to clone a repository. It works pretty well for small repositories, but seems to fail on larger repositories and I do not understand why.
All that my code does is send a hard-coded "want" request. As I understand it, responses come with a four-byte size prefix, then the data of the size - 4. The first four bytes are the size of the object in ASCII and after that size, a new object should begin by specifying its size. The final "0000" string should signify the end. On small repositories my code works fine. But when use a large git repository, I hit an object size that cannot be interpreted by ASCII into a number. In particular, right before the crash I am pretty consistently getting a size starting with 0x32,0x00 or 0x32,0x30,0x00 (note, it starts with 0x32 and has 0x00 in there). This is not a readable four byte ascii value, likely an erroneous size value, which causes the next object size calculation to land incorrectly and subsequently the program to malfunction. My questions: 1. Am I misunderstanding the protocol? 2. Does 0x32 signify something? 3. Also, where can I see in the source code git parses the data it receives from a "want" request? If you would like to see my code, it is located here: http://git.farhan.codes/farhan/post To compile to Linux run: gcc post.c main.c -lcurl -o post To compile on FreeBSD you can use the Makefile or: cc -L /usr/local/lib -I /usr/local/include -lcurl main.c post.c -o post In both cases you need to have libcurl installed. To run do: ./post [a git http url] [a commit value] ie: ./post http://git.farhan.codes/farhan/openbsd 373dd5ba116d42cddf1fdcb58b578a4274f6d589 I have a lot of debugging printf() messages, but it eventually hits a point where you see this: ========Start========= Current stats: Current state: 999 Received: 1448 We have enough bytes, moving forward == New Object No Error, object is 32 bytes Size bytes: 32 30 00 00 The program interprets the size of {0x32,0x30,0x00,0x00} to be "20" which in decimal is 32, causing the next read to fail. This problem repeats on a few different repositories. Any assistance is welcome, I am very stuck on how the HTTP git protocol works. Thanks, -- Farhan Khan PGP Fingerprint: B28D 2726 E2BC A97E 3854 5ABE 9A9F 00BC D525 16EE
