Premailer inline css ordering is different across machines

[cpoo22]

When upgrading premailer-rails from 1.9.2 (premailer 1.8.6) to 1.10.2 (premailer 1.11.1) the order of the inlined styles changed on different machines.

For a regression spec that creates an email and then checks the HTML part on my local (MAC) I have

<td class="font-light font-12 footer" bgcolor="#262626" style="color: #FFFFFF; padding: 20px 30px; font-family: museo-sans-rounded, helvetica, arial; font-size: 12px; line-height: 1.67; border-top-width: 7px; border-top-color: #6F58A2; border-top-style: solid; -webkit-text-size-adjust: 100%; -ms-text-size-adjust: 100%; mso-table-lspace: 0pt; mso-table-rspace: 0pt;">

But when the spec is run on semaphore I get a failure with

<td class="font-light font-12 footer" bgcolor="#262626" style="color: #FFFFFF; font-family: museo-sans-rounded, helvetica, arial; font-size: 12px; line-height: 1.67; border-top-width: 7px; border-top-color: #6F58A2; border-top-style: solid; -webkit-text-size-adjust: 100%; -ms-text-size-adjust: 100%; mso-table-lspace: 0pt; mso-table-rspace: 0pt; padding: 20px 30px;">

You can see that the padding has moved to the end. This random ordering is making writing regression test difficult.

[meowsus]

Though I'm not sure what has caused the ordering to change, I'm not sure that this is an actual bug with the software. The order of CSS in an on-line style attribute shouldn't have an effect on the rendered email, visually speaking, unless, of course, you have duplicate properties being output.

Since this issue is old and since it should be possible to write a test that does not assert the order of the style attribute value in your project, I'm going to close this.

Feel free to reopen if you disagree :)

[meowsus]

Though I'm not sure what has caused the ordering to change, I'm not sure that this is an actual bug with the software. The order of CSS in an on-line style attribute shouldn't have an effect on the rendered email, visually speaking, unless, of course, you have duplicate properties being output.

Since this issue is old and since it should be possible to write a test that does not assert the order of the style attribute value in your project, I'm going to close this.

Feel free to reopen if you disagree :)

[foxyblocks]

I've encountered this issue. We have tests that spit out html results of emails that we commit to the repository. This allows us to see previews in source control as well as diff them during CI to catch un-intended regressions. They are breaking due to the order change.

It may be that we shouldn't be testing this but I think it's also true that we'd want html rendering to be predictable.

Is there a possible fix?

[meowsus]

@foxyblocks I'm sure that there is a possible fix. I'll reopen this issue, but don't have the bandwidth right now to take it on. If you have some time, would you be able to look into it? Even writing a failing test that showcases the bug would be useful in triaging.