Base64 base64decode does not ignore whitespace, but should

[christopher-dG]

I'm not sure that there's actually anything in the RFC about this, but base64 strings often have newlines at regular intervals (GitHub for example sends file contents as 60-char lines).

julia> using Base64

julia> s = "bmFtZSA9ICJHZW5pZSIKdXVpZCA9ICJjNDNjNzM2ZS1hMmQxLTExZTgtMTYx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\nZGYtYWIzZjljZjk5ZWI2IgoKW2NvbXBhdF0KanVsaWEgPSAiMSIK"

julia> println(String(base64decode(s)))
name = "Genie"
uuid = "c43c736e-a2d1-11e8-161f-af95117fbd1e"
authors = ["Adrian Salceanu <e@essenciary.com>"]
version = "0.9.4"

[deps]
ArgParse = "c7e460c6-2fb9-53a9-8c5b-16f535851c63"
Dates = "ade2ca70-3891-5945-98fb-dc099432e06a"
Distributed = "8ba89e20-285c-5b6f-9357-94700520ee1b"
Gumbo = "708ec375-b3d6-5a57-a7ce-8257bf98657a"
HTTP = "cd3eb016-35fb-5094-929b-558a96fad6f3"
HttpCommon = "77172c1b-203f-54ac-aa54-3f1198fe9f90"
JSON = "682c06a0-de6a-54ab-a142-c8b1cf79cde6"
Markdown = "d6f4376e-aef5-505a-96c1-9c027394607a"
MbedTLS = "739be429-bea8-5141-9913-cc70e7f3736d"
Millboard = "39ec1447-df44-5f4c-beaa-866f30b4d3b2"
MiniLogging = "f59402ec-0262-5707-a561-770af94bc5a6"
Nettle = "49dea1ee-f6fa-5aa6-9a11-8816cee7d4b9"
Nullables = "4d1e1d77-625e-5b40-9113-a560ec7a8ecd"
OrderedCollections = "bac558e1-5e72-5ebc-8fee-abe8a469f55d"
Pkg = "44cfe95a-1eb2-52ea-b672-e2afdf69b78f"
Reexport = "189a3867-3050-52da-a836-e630ba90ab69"
Revise = "295af30f-e4ad-537b-8983-00126c2a3abe"
SHA = "ea8e919c-243c-51af-8825-aaa63cd721ce"
Serialization = "9e88b42a-f829-5b0c-bbe9-9e923198166b"
Sockets = "6462fe0b-24de-5631-8697-dd941f90decc"
URIÄÄòser = "30578b45-9adc-5946-b283-645ec420af67"
Unicode = "4ec0a83e-493e-50e2-b9ac-8f72acf5a8f5"
YAML = "ddb6d928-2868-570f-bddf-ab3f9cf99eb6"

[compat]
julia = "1"

URIÄÄòser is a pretty interesting way to spell URIParser.

Plugging that same string into Python's base64.b64decode works as expected.
Applying replace(s, r"\s" => "") before decoding works, but it shouldn't be necessary and probably performs less than optimally.

Discovered through debugging this.

[Moelf]

according to this post and the Wikipedia page it points to, \s characters in the string should be ignored when decoding, will try to make a PR

[christopher-dG]

It's not fixed, the example above has the same issue on 1.1 and master.

[Moelf]

Why is that I cannot reproduce this?

julia> using Base64
julia> s = "bmFtZSA9ICJHZW5pZSIKdXVpZCA9ICJjNDNjNzM2ZS1hMmQxLTExZTgtMTYx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\nZGYtYWIzZjljZjk5ZWI2IgoKW2NvbXBhdF0KanVsaWEgPSAiMSIK"
"bmFtZSA9ICJHZW5pZSIKdXVpZCA9ICJjNDNjNzM2ZS1hMmQxLTExZTgtMTYx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\nZGYtYWIzZjljZjk5ZWI2IgoKW2NvbXBhdF0KanVsaWEgPSAiMSIK"

julia> println(String(base64decode(s)))
name = "Genie"uuid = "c43c736e-a2d1-11e8-161f-af95117fbd1e"authors = ["Adrian Salceanu <e@essenciary.com>"]version = "0.9.4"[deps]ArgParse = "c7e460c6-2fb9-53a9-8c5b-16f535851c63"Dates = "ade2ca70-3891-5945-98fb-dc099432e06a"Distributed = "8ba89e20-285c-5b6f-9357-94700520ee1b"
Gumbo = "708ec375-b3d6-5a57-a7ce-8257bf98657a"
HTTP = "cd3eb016-35fb-5094-929b-558a96fad6f3"
HttpCommon = "77172c1b-203f-54ac-aa54-3f1198fe9f90"
JSON = "682c06a0-de6a-54ab-a142-c8b1cf79cde6"
Markdown = "d6f4376e-aef5-505a-96c1-9c027394607a"
MbedTLS = "739be429-bea8-5141-9913-cc70e7f3736d"
Millboard = "39ec1447-df44-5f4c-beaa-866f30b4d3b2"
MiniLogging = "f59402ec-0262-5707-a561-770af94bc5a6"
Nettle = "49dea1ee-f6fa-5aa6-9a11-8816cee7d4b9"
Nullables = "4d1e1d77-625e-5b40-9113-a560ec7a8ecd"
OrderedCollections = "bac558e1-5e72-5ebc-8fee-abe8a469f55d"
Pkg = "44cfe95a-1eb2-52ea-b672-e2afdf69b78f"
Reexport = "189a3867-3050-52da-a836-e630ba90ab69"
Revise = "295af30f-e4ad-537b-8983-00126c2a3abe"
SHA = "ea8e919c-243c-51af-8825-aaa63cd721ce"
Serialization = "9e88b42a-f829-5b0c-bbe9-9e923198166b"
Sockets = "6462fe0b-24de-5631-8697-dd941f90decc"
URIser = "30578b45-9adc-5946-b283-645ec420af67"
Unicode = "4ec0a83e-493e-50e2-b9ac-8f72acf5a8f5"
YAML = "ddb6d928-2868-570f-bddf-ab3f9cf99eb6"

[compat]
julia = "1"

[christopher-dG]

Well that sure is different, but it isn't correct either.

[Moelf]

Also, I find this behavior non local:

julia> s = "MWViMi01MmVhLWI2NzItZTJhZmRmNjliNzhmIgpSZWV4cG9ydCA9ICIxODlh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\nZGYtYWIzZjljZjk5ZWI2IgoKW2NvbXBhdF0KanVsaWEgPSAiMSIK";

julia> using Base64

julia> println(String(Base64.base64decode(s)))
1eb2-52ea-b672-e2afdf69b78f"
Reexport = "189a3867-3050-52da-a836-e630ba90ab69"
Revise = "295af30f-e4ad-537b-8983-00126c2a3abe"
SHA = "ea8e919c-243c-51af-8825-aaa63cd721ce"
Serialization = "9e88b42a-f829-5b0c-bbe9-9e923198166b"
Sockets = "6462fe0b-24de-5631-8697-dd941f90decc"
URIParser = "30578b45-9adc-5946-b283-645ec420af67"
Unicode = "4ec0a83e-493e-50e2-b9ac-8f72acf5a8f5"
YAML = "ddb6d928-2868-570f-bddf-ab3f9cf99eb6"

[compat]
julia = "1"

@bicycle1885

[bicycle1885]

@Moelf Sorry, I don't understand your point. Anything wrong in your example? I think it seems to be working as I expect.

~ $ echo "
    MWViMi01MmVhLWI2NzItZTJhZmRmNjliNzhmIgpSZWV4cG9ydCA9ICIxODlh
    Mzg2Ny0zMDUwLTUyZGEtYTgzNi1lNjMwYmE5MGFiNjkiClJldmlzZSA9ICIy
    OTVhZjMwZi1lNGFkLTUzN2ItODk4My0wMDEyNmMyYTNhYmUiClNIQSA9ICJl
    YThlOTE5Yy0yNDNjLTUxYWYtODgyNS1hYWE2M2NkNzIxY2UiClNlcmlhbGl6
    YXRpb24gPSAiOWU4OGI0MmEtZjgyOS01YjBjLWJiZTktOWU5MjMxOTgxNjZi
    IgpTb2NrZXRzID0gIjY0NjJmZTBiLTI0ZGUtNTYzMS04Njk3LWRkOTQxZjkw
    ZGVjYyIKVVJJUGFyc2VyID0gIjMwNTc4YjQ1LTlhZGMtNTk0Ni1iMjgzLTY0
    NWVjNDIwYWY2NyIKVW5pY29kZSA9ICI0ZWMwYTgzZS00OTNlLTUwZTItYjlh
    Yy04ZjcyYWNmNWE4ZjUiCllBTUwgPSAiZGRiNmQ5MjgtMjg2OC01NzBmLWJk
    ZGYtYWIzZjljZjk5ZWI2IgoKW2NvbXBhdF0KanVsaWEgPSAiMSIK
    " | base64 --decode
1eb2-52ea-b672-e2afdf69b78f"
Reexport = "189a3867-3050-52da-a836-e630ba90ab69"
Revise = "295af30f-e4ad-537b-8983-00126c2a3abe"
SHA = "ea8e919c-243c-51af-8825-aaa63cd721ce"
Serialization = "9e88b42a-f829-5b0c-bbe9-9e923198166b"
Sockets = "6462fe0b-24de-5631-8697-dd941f90decc"
URIParser = "30578b45-9adc-5946-b283-645ec420af67"
Unicode = "4ec0a83e-493e-50e2-b9ac-8f72acf5a8f5"
YAML = "ddb6d928-2868-570f-bddf-ab3f9cf99eb6"

[compat]
julia = "1"

[Moelf]

@bicycle1885 That's my point. The bug only appears when you have the complete s string from the beginning of the post. Even though bug happens on the line of URIParser, only use those lines from the encoded string will not reproduce the bug, thus non-local.

[bicycle1885]

Ah, okay. It makes sense.

[jsulmont]

Hi -- I've submitted a #32457 which fixes the issue.
I have yet to come up with a test (possibly shorter than the one provided by @christopher-dG).
The title for this issue is probably misleading, as base64decode() correctly ignores whichever UInt8 input isn't in the encoding table. The bug is triggered when read_to_buffer is called with a non empty buffer (and whatever hasn't been consumed yet has to be copied at the beginning of the buffer).