XHTTP client: Fix edge-case issue for packet-up mode

[Fangliding]

今天刚好聊到了(

[patterniha]

I fix that. also because "return 0, err" may not correct on your code(we may send some data and then get error)

[Fangliding]

能不能别乱动 这样不行 会出问题的

[patterniha]

Can you please stop moving around? It will cause problems.

although it has no practical effect,
but we should return total-success-send-bytes, err instead of return 0, err.

[Fangliding]

Can you please stop moving around? It will cause problems.

although it has no practical effect, but we should return total-success-send-bytes, err instead of return 0, err.

首先这不是我写的

第一个buffer.Write 那是写buffer 预处理 并没有东西写入底层连接
第二个WriteMultiBuffer 只给了一个err

[patterniha]

i talk about WriteMultiBuffer,

suppose after buffer.Write(b), buffer is equal to [b0, b1].

and in for loop WriteMultiBuffer(buf.MultiBuffer{b0}) return no error,
but WriteMultiBuffer(buf.MultiBuffer{b1}) return error.

so we should return b0.Len(), err instead of 0, err, even if it has no practical effect.

[patterniha]

thx.

[RPRX]

我看改后似乎没啥问题，可以合了吗

[Fangliding]

没啥问题 第一个commit就是好的 后面都无关紧要

[patterniha]

I think there is no problem after the modification. Can we merge it?

instead of number of bytes-sent, it always return 0 on error, it is OK now.

this PR, and #5006 fixes important issues and should be merged for next version.

then I release serverless-for-iran-anti-sanction.

I hope I don't find any new bugs while testing this time, It was really strange, as if someone had intentionally put some hidden bugs in Xray-core.

[RPRX]

感觉你的 serverless-for-iran-anti-sanction 像我的 REALITY 文章、Switch、GUI 客户端、Rust core 等大饼一样遥遥无期

[patterniha]

I worked hard for it, I hope it will be well received this time.

At least it caused many bugs to be fixed in Xray-core :)

[Fangliding]

没看出来 fix 了个啥important 绑在你那个pr催合并吗 还跑到这底下提一嘴你的玩意 喊得到大声。。

[RPRX]

你们不要再打了啦 这样是打不死人的

怕非中文母语者不懂：玩梗，无恶意

[RPRX]

我看得出来这个 PR 修了个 bug，但我想梳理清 read/write 间的关系，半年没写 XHTTP 竟一时看不懂了卧槽，当时没留太多注释

我算是明白了，原来注释主要是给半年后的自己看的，方便快速恢复记忆

[patterniha]

I have fulfilled all your requests without any expectations until now.

I have a goal and I fight for it. No harash word can upset me. at least I think @RPRX appreciated my efforts like a father, thank you very much again.

[RPRX]

不需要为无意义的东西争吵，@patterniha 你那个 PR 我等下也看看吧

[patterniha]

No need to argue over meaningless things.@patternihaI'll take a look at your PR later.

i hope before release next-version.

[RPRX]

不过谈到 Write() 的 err，在 TCP 上返回 n 没有意义，因为都 err 了你也不能确定对端收了多少，也不敢重新 Write()，而是会放弃整条连接，所以其实返回 0 就够了（除非像 buffer 这些内部函数），我写的 encryption 就全返回 0，Read() 的 err 的话 n 还有点意义

[patterniha]

I also said it has no practical effect, I was just saying it in terms of programming logic: #5020 (comment)

Also, the main problem was something else, and this was just a side issue.

[RPRX]

不过在原生 UDP 上 Write() err 时返回 n 更没啥意义，操作系统估计都只会返回 0，不同之处只在于 UDP 上敢重新 Write()

[patterniha]

But for debugging and correct-calculation-of-traffic-statistics we must return exact value.